Fact-checked by Grok 2 weeks ago

Affective computing

Affective computing is computing that relates to, arises from, or deliberately influences emotions or other affective phenomena. Coined by in her seminal paper, the field focuses on enabling machines to recognize, interpret, and respond to human emotional states to facilitate more natural human-computer interactions. It draws from interdisciplinary foundations in , , , and to bridge the gap between emotional human experiences and technological systems. Picard expanded on these ideas in her 1997 book Affective Computing, published by the , which provided the intellectual framework for the discipline and emphasized the role of emotions in rational decision-making and perception, as supported by neurological research such as Antonio Damasio's work on emotion and reason. The field originated at the , where Picard's Affective Computing Research Group continues to pioneer advancements in Emotion AI—a core subset involving the detection and simulation of emotions through computational models. Over the past three decades, affective computing has evolved from theoretical models of to practical implementations, incorporating techniques for analyzing multimodal data like facial expressions, vocal tones, physiological signals (e.g., ), and body gestures; recent advancements include integration with large language models for enhanced multimodal emotion understanding. Key components of affective computing include , which uses sensors and algorithms to detect affective states; affective expression, enabling systems to convey emotions via synthetic speech, avatars, or adaptive interfaces; and affective influence, where technology modulates user emotions for beneficial outcomes. These elements are integrated into wearable devices and software to advance theory and research by collecting real-world data on affective responses. Notable applications span multiple domains, including monitoring to forecast and prevent conditions like through passive tracking; educational tools that adapt content based on learner or ; human-robot for empathetic companionship; and workplace systems for detection and enhancement. In healthcare, affective technologies support interventions by modeling emotional development and aid communication for individuals with expressive challenges. Recent developments emphasize ethical considerations, such as data in physiological sensing and explainable for transparent inference, ensuring responsible deployment amid growing integration with .

Introduction and History

Definition and Scope

Affective computing is a branch of that enables machines to recognize, interpret, process, and simulate human emotions, a concept first coined by in her 1995 technical report Affective Computing and expanded upon in her 1997 book of the same name. In this foundational work, Picard describes affective computing as systems designed to relate to, arise from, or deliberately influence emotions, emphasizing the need for computers to interact more naturally with humans by incorporating emotional awareness. This field emerged from the recognition that emotions play a critical role in human , , and social interaction, extending beyond traditional rational models of intelligence. The scope of affective computing is inherently multidisciplinary, integrating principles from , , to build emotionally . It focuses on bridging the gap between human affective experiences and machine capabilities, allowing for more empathetic and adaptive technologies in areas like human-computer interaction. Central to this scope are the core components of the field: affect detection, which identifies emotional states through sensory inputs; affect interpretation, which assigns contextual meaning to detected emotions; and affect synthesis, which enables machines to generate and express appropriate emotional responses. These elements form the backbone of systems that can perceive and respond to human affects in . A key aspect of affective computing involves modeling human emotions, often drawing on established psychological theories to inform computational approaches. Categorical models, such as Paul Ekman's framework of six basic emotions—happiness, sadness, fear, anger, surprise, and disgust—treat emotions as discrete universal categories identifiable across cultures. In contrast, dimensional models like James Russell's 1980 circumplex model represent emotions on a two-dimensional plane of (pleasantness-unpleasantness) and (activation level), providing a continuous spectrum for more nuanced representation. A prerequisite for advancing in affective computing is a solid grasp of these human emotion theories, as they underpin the design of reliable detection and simulation techniques without which technical implementations lack psychological validity.

Historical Development and Key Figures

The roots of affective computing trace back to the 1980s in research, where scholars began exploring the role of in and . Marvin Minsky's 1986 book The Society of Mind argued that are essential components of intelligent behavior, emerging from interactions among simpler cognitive processes, and influenced subsequent work on integrating into . This perspective laid theoretical groundwork by challenging the prevailing view of intelligence as purely rational, emphasizing how guide , learning, and in complex environments. The field was formally established in the mid-1990s through Rosalind Picard's pioneering contributions at the . In her seminal 1995 technical report "Affective Computing," Picard defined the discipline as computing that relates to, arises from, or deliberately influences emotions, proposing models for machines to recognize and express affect using physiological and behavioral cues. This was expanded in her 1997 book Affective Computing, which advocated for wearable sensors to monitor emotions in and founded the Affective Computing Group in 1997 to advance interdisciplinary research. Key figures emerged alongside these developments: , who in the late 1990s developed , an expressive robot head at that demonstrated affective interaction through expressions and , pioneering emotion in social . Björn Schuller advanced speech-based from the early 2000s, contributing foundational methods for acoustic feature analysis and linguistic integration in hybrid models. Major milestones marked the field's evolution through the 2000s and . In the 2000s, integration of the (FACS) enabled automated analysis for detection, as seen in early systems that coded action units for robust recognition. The saw the rise of multimodal fusion techniques, combining modalities like speech, face, and physiology to improve accuracy, with reviews highlighting feature-level and decision-level approaches for hybrid inference. Institutional efforts supported standardization, including the founding of the HUMAINE Association in 2005 (now the Association for the Advancement of Affective Computing), which organized the first International Conference on Affective Computing and Intelligent Interaction to foster global collaboration. Post-2020 advancements integrated and large language models, enabling real-time AI via transformer architectures for recognition, as evidenced in challenges like MER 2025 exploring forecasting with pre-trained models. These developments, building on Picard's vision, have scaled affective systems for diverse applications while addressing ethical considerations in -aware computing.

Core Concepts

Emotion Detection and Recognition

Emotion detection and recognition in affective computing involves a systematic process beginning with the sensing of raw data from human users, such as audio signals, visual cues, or physiological measurements, to capture indicators of emotional states. This raw data undergoes feature extraction, where relevant attributes are identified and quantified—for instance, prosodic elements like pitch variation in speech or action units in facial expressions—to represent emotional content in a computationally tractable form. The extracted features are then fed into classification algorithms, which map them to either discrete emotion categories (e.g., , , anger) based on categorical models or continuous dimensions such as (positive-negative) and (high-low intensity) using dimensional frameworks. Theoretical foundations for emotion detection draw heavily from psychological models that emphasize cognitive evaluation of stimuli. posits that emotions arise from an individual's subjective assessment of events in relation to personal goals and , where primary appraisal evaluates the relevance and of a stimulus, and secondary appraisal assesses potential. This framework informs computational models by guiding toward indicators of evaluative processes, such as changes in physiological signaling threat relevance. Complementing this, Scherer's component process model (1984) conceptualizes emotions as dynamic, emergent episodes resulting from synchronized changes across multiple subsystems: of novelty and goal conduciveness, autonomic physiological responses, motivational action tendencies, motor expressions, and subjective feelings. In affective computing, this model supports by modeling the temporal sequencing of these components, enabling systems to infer emotions from patterns of rather than isolated signals. Unimodal detection, relying on a single input like speech, typically achieves accuracies of 60-70% in controlled settings due to limitations in capturing the full spectrum of emotional cues, such as contextual ambiguities in prosody alone. In contrast, approaches integrate data from multiple sources (e.g., speech, expressions, and gestures), yielding substantial improvements through fusion, with meta-analyses reporting an average 8.12% gain over the best unimodal method and accuracies reaching up to 90% in laboratory environments where is optimized. These benefits arise from complementary information across modalities, reducing errors from noise or modality-specific variability, though real-world deployment faces challenges like asynchronous inputs. Performance in emotion recognition is evaluated using standard machine learning metrics to assess reliability across diverse emotional classes. Accuracy measures the overall proportion of correct predictions, while precision quantifies the fraction of positive identifications that are truly positive, and recall captures the fraction of actual positives correctly identified; the F1-score, as their harmonic mean, balances these for imbalanced datasets common in emotion tasks. Confusion matrices further visualize misclassifications, highlighting pairwise errors such as conflating surprise with fear due to overlapping arousal patterns. These metrics are essential for benchmarking, as high accuracy alone may mask poor recall for subtle emotions like disgust. Context plays a critical role in emotion detection, as situational and cultural factors modulate expression through display rules—social norms dictating when and how emotions are shown or suppressed. For example, cultures emphasizing collectivism, such as , often enforce stronger rules for masking negative emotions in public to maintain harmony, leading to subdued expressions that unimodal systems trained on Western data may misinterpret as neutral. Incorporating contextual priors, like cultural display norms, enhances robustness by adjusting thresholds for variability in emotional and across groups.

Emotion Simulation and Expression in Machines

Emotion simulation in machines involves computational methods to generate internal emotional states and express them through various output channels, enabling more natural human-machine interactions. Early approaches relied on rule-based systems, where predefined if-then rules map situational inputs to emotional responses, such as triggering an response when detecting in a conversational . These systems, inspired by psychological theories, provide deterministic and interpretable simulations but lack flexibility for complex, context-dependent . In contrast, modern generative models create dynamic emotional expressions by learning from data to produce realistic outputs, such as animating facial features to convey subtle or from neutral inputs; recent advances as of 2025 incorporate large language models (LLMs) for generating emotionally-aligned responses in dialogue systems, shifting toward generative paradigms beyond traditional categorical frameworks. Expression modalities in affective computing encompass multiple channels to convey simulated effectively. Virtual agents often use dynamic animations, where machine-generated expressions mimic micro-expressions to build , as seen in embodied conversational agents that adjust raises or smiles based on appraised emotional intensity. Tonal integrates prosodic features like pitch variation and to infuse spoken responses with affective , allowing systems to compassionate during user distress. Haptic feedback serves as a tactile , using vibrations or patterns on wearable devices to transmit , such as rhythmic pulses simulating warmth for or irregular jolts for anxiety, enhancing immersion in environments. Computational models underpin these simulations by formalizing how machines appraise and generate emotions. The OCC model, originally a psychological framework, has been adapted for computational appraisal, where agents evaluate events relative to goals, standards, and tastes to derive emotions like or reproach, enabling rule-based or probabilistic simulations in virtual characters. The EMA (Emotion and Adaptation) architecture extends this by modeling dynamic appraisal processes over time, incorporating coping strategies to produce believable emotional expressions in agents, such as shifting from to acceptance in response to unfolding scenarios. These models prioritize cognitive structures to ensure simulated emotions align with human-like reasoning, facilitating applications in interactive systems. Evaluation of emotion simulation focuses on perceived authenticity and impact rather than internal accuracy. Variants of the Turing Test assess emotional believability by having users distinguish machine-generated responses from human ones in empathetic dialogues, revealing high indistinguishability in advanced systems. User studies measure perceived empathy through scales like the Perceived Empathy of Technology Scale (PETS), which evaluates factors such as emotional responsiveness and trust, showing that expressive virtual agents increase user satisfaction in interaction tasks. These metrics emphasize subjective human judgments to refine simulation techniques. An ethical consideration in emotion simulation is the risk of , where users over-attribute genuine feelings to machines, potentially leading to emotional or . Studies indicate that intentionally harming emotion-expressing robots heightens perceptions of their and status, blurring boundaries and raising concerns about psychological harm or misguided trust in non-sentient systems. Designers must balance expressiveness with to mitigate while preserving interaction benefits.

Sensing Technologies

Speech-Based Emotion Recognition

Speech-based emotion recognition involves analyzing audio signals to detect emotional states conveyed through vocal expressions, focusing on paralinguistic elements that transcend linguistic content. This approach leverages the acoustic properties of speech, such as variations in and , to infer emotions like , , or . Key to this process is the extraction of acoustic features that capture the nuances of emotional vocalization. Prosodic features, including (fundamental frequency), (speech rate), and volume (energy or ), provide temporal and dynamic indicators of emotion; for instance, elevated and faster often signal excitement or . Spectral descriptors further enhance recognition by modeling the frequency content of speech. Among these, Mel-Frequency Cepstral Coefficients (MFCCs) are widely used, representing the short-term power spectrum of sound on a nonlinear that approximates human auditory . The MFCCs are computed as: c_n = \sum_{k=1}^K \log(S_k) \cos\left(\frac{\pi n (k-0.5)}{K}\right), where S_k are the outputs of mel-scale filters applied to the signal's power spectrum, K is the number of filters, and n indexes the coefficients. These features effectively distinguish emotional categories by highlighting formant structures and harmonic variations in voiced segments. Algorithms for speech-based have evolved from traditional statistical models to architectures tailored for audio processing. Hidden Markov Models (HMMs) were seminal in early systems, modeling the sequential nature of speech prosody to classify emotions through state transitions based on feature sequences like MFCCs and pitch contours. Convolutional Neural Networks (CNNs) advanced this by treating spectrograms—time-frequency representations of audio—as images, applying filters to detect local patterns indicative of emotional or . In the 2020s, transformer-based models, such as wav2vec 2.0, enabled end-to-end by learning contextual representations from raw waveforms via self-supervised pretraining on large speech corpora, followed by on emotion-labeled data. These transformers capture long-range dependencies in audio, outperforming prior methods on dynamic emotional expressions. Recent advances as of 2025 include adaptations of models like and Whisper for SER, supporting naturalistic speech in challenges such as Interspeech 2025, with improved robustness to noise and dialects. Despite progress, speech-based emotion recognition faces significant challenges, including speaker variability, where individual differences in voice quality and degrade model generalization across users. Noise robustness remains a critical issue, as environmental can mask subtle prosodic cues, necessitating robust or denoising preprocessing. Cultural differences in vocal emotion expression further complicate deployment; for example, anger may manifest with higher in Western speakers but lower in some Asian cultural contexts, leading to cross-cultural misclassifications. These factors underscore the need for diverse, inclusive training data to mitigate biases. Performance benchmarks on datasets like IEMOCAP, which includes interactions with acted and improvised emotions, typically yield 65-75% accuracy for four-class (e.g., angry, happy, sad, ), with weighted accuracies around 72-74% accounting for class imbalance. These results highlight the modality's potential but also its limitations compared to human-level , particularly for subtle or mixed emotions. To enhance accuracy, paralinguistic analysis often integrates speech features with textual sentiment from transcribed words, combining prosodic indicators with lexical cues like positive or negative phrasing. This fusion, typically via classifiers, improves overall detection by resolving ambiguities where vocal contradicts verbal content, such as sarcastic remarks.

Facial Expression Analysis

Facial expression analysis is a cornerstone of affective computing, focusing on the automated interpretation of human emotions through visual cues from facial movements and configurations. This involves detecting and classifying expressions to infer emotional states, enabling machines to respond empathetically in human-computer interactions. Techniques in this domain process both static images and dynamic video sequences, emphasizing the subtlety of facial to distinguish between basic emotions such as , , , , , and . Feature extraction forms the initial step in facial expression analysis, where key facial components are identified to capture emotional indicators. Landmark detection, for instance, locates specific points on the face, with the widely adopted 68-point model delineating contours around the eyes, eyebrows, nose, mouth, and jawline to quantify deformations associated with expressions. This model, implemented in libraries like , facilitates precise tracking of facial geometry for emotion . Additionally, methods analyze pixel motion between consecutive frames to detect micro-expressions—brief, involuntary facial movements lasting less than 1/25 of a second that reveal concealed emotions; these are particularly useful for applications in detection and monitoring. The (FACS), developed by and Wallace V. Friesen in 1978, provides a foundational framework for dissecting facial expressions into atomic components known as Action Units (AUs). FACS anatomically maps 44 AUs to specific muscle actions, such as AU12 (lip corner puller), which activates the zygomatic major muscle to produce a indicative of . Complex emotions arise from AU combinations; for example, is characterized by AU1 (inner brow raiser) + AU2 (outer brow raiser) + AU5 (upper lid raiser), resulting in widened eyes and raised eyebrows. This system enables systematic annotation and has been certified for reliability in over 3,000 studies, influencing both manual coding and automated tools in affective computing. Classification methods in facial expression analysis leverage machine learning to map extracted features to emotional categories, contrasting AU-based approaches—which decompose expressions into independent muscle activations for granular analysis—with holistic methods that treat the face as a unified for direct prediction. Early techniques employed Support Vector Machines (SVMs) on handcrafted features like Gabor wavelets, achieving up to 88% accuracy on posed expressions in controlled settings. has advanced this field, with convolutional neural networks (CNNs) like ResNet applied to datasets such as FER2013—a comprising 35,887 grayscale images of varied expressions—yielding approximately 70-75% accuracy on test sets, though performance varies by due to class imbalance. AU-based classifiers often outperform holistic ones in spontaneous scenarios by isolating subtle cues, as demonstrated in comparative studies where component-wise analysis improved recognition by 10-15% over whole-face processing. Despite progress, analysis faces significant challenges, including occlusions from masks or hands, head pose variations that distort feature alignment, and the disparity between posed (deliberate, exaggerated) and spontaneous (natural, subtle) expressions. Spontaneous expressions, which better reflect authentic , exhibit different muscle activation patterns and dynamics compared to posed ones, leading to substantially lower recognition accuracies—often around 50-60% in real-world settings versus 80-90% for posed data in labs. These issues are exacerbated in unconstrained environments, where lighting, , and cultural differences further degrade performance. Recent advances, particularly from 2024-2025, emphasize real-time facial expression analysis via for mobile devices, enabling low-latency, privacy-preserving detection without cloud dependency. Lightweight architectures like MobileNet and EfficientNet, optimized for resource-constrained hardware, achieve over 70% accuracy in on-device inference while processing video at 30 , supporting applications in wearable tech and . These developments integrate FACS-inspired AU detection with transformer-based models for robust handling of variations, marking a shift toward deployable affective interfaces.

Physiological Signal Monitoring

Physiological signal monitoring in affective computing involves capturing involuntary bodily responses to infer emotional states, providing a covert and objective measure of internal and that complements visible cues. These signals, primarily from the , reflect reactions such as increased sweating or fluctuations during emotional experiences. Unlike explicit expressions, physiological monitoring enables real-time, non-invasive detection in naturalistic settings, with applications in and human-machine interfaces. Electrodermal activity (EDA), also known as galvanic skin response (GSR), measures changes in conductance due to activity, serving as a key indicator of emotional . EDA signals increase with activation, correlating strongly with high- states like excitement or , while showing weaker links to . conductance G is calculated as G = \frac{I}{V}, where I is the current and V is the applied voltage, allowing quantification of (baseline) and phasic (event-related) components. Meta-analyses confirm EDA's superior performance in prediction over , with accuracies often exceeding 70% in dimensional models. Heart rate variability (HRV), derived from electrocardiogram (ECG) signals, assesses fluctuations in inter-beat intervals to detect emotional stress and . HRV decreases under stress due to parasympathetic withdrawal and sympathetic dominance, with revealing key metrics like the low-frequency (LF) to high-frequency (HF) power ratio (LF/HF), where elevated ratios indicate heightened stress. For instance, meta-analyses of 37 studies show consistent reductions in HF power and increases in LF during acute stress, linking HRV to activity for appraisal. This makes HRV a reliable for distinguishing calm from anxious states in affective systems. Facial electromyography (EMG) captures subtle muscle activations to gauge emotional , focusing on non-visible contractions. The , involved in smiling, activates during positive emotions like , reflecting approach-oriented . Conversely, the , associated with frowning, engages during negative states such as or , indicating withdrawal tendencies. Studies demonstrate reliable differentiation, with zygomaticus activity rising to happy stimuli and corrugator to aversive ones, enabling accuracies around 80% in controlled settings. Blood volume pulse (BVP), measured via photoplethysmography (PPG), tracks peripheral blood flow changes to identify emotional , particularly . PPG sensors detect variations, with acceleration features (e.g., second derivative) showing abrupt shifts during responses due to . In datasets eliciting emotions like "scary," BVP features achieve up to 71.88% recognition accuracy using , highlighting time-frequency domain metrics like for distinguishing high- negatives. This approach suits wearable integration for real-time detection in safety-critical applications. Facial color analysis uses RGB imaging to detect chromaticity shifts signaling emotions, bypassing overt expressions. , indicated by increased redness from , conveys in states like or , while (paleness) from signals or . Remote PPG (rPPG) enhances this by extracting pulse from color variations, with experiments showing 70% accuracy in decoding 18 emotions from color alone and 85% for . These patterns, driven by blood flow, provide an efficient, for emotional transmission. Recent developments as of emphasize fusion of physiological signals, such as EEG and ECG, using to boost recognition accuracy in real-world scenarios, achieving up to 95% in controlled elicitation via . Wearable devices like smartwatches integrate GSR and HRV sensors for continuous physiological monitoring, enabling affective computing. For example, wrist-based EDA and ECG capture in daily life, with preprocessing mitigating issues. However, motion artifacts from user movement distort signals, particularly in PPG and EDA, requiring adaptive filtering for reliability. concerns also arise from persistent , necessitating secure protocols to protect sensitive emotional insights.

Gesture and Body Language Recognition

Gesture and body language recognition in affective computing involves the analysis of non-verbal cues such as postures, movements, and dynamic gestures to infer emotional states, providing a complementary modality to facial or vocal signals. This approach leverages computer vision techniques to detect and interpret body dynamics, enabling machines to understand human affect through skeletal structures and motion patterns. Key features extracted include pose estimation keypoints, which represent joint positions across the body, and kinematic attributes like limb velocity and acceleration to capture agitation or fluidity in movements. For instance, OpenPose, a widely adopted real-time pose estimation library, detects 25 keypoints for the human body (including head, torso, limbs, and feet), facilitating the modeling of full-body configurations for emotion inference. Emotions are mapped to specific gesture patterns based on psychological and computational models; for example, open postures often signal welcoming or , while crossed arms indicate defensiveness associated with or discomfort. These mappings draw from established research, where expansive s correlate with positive affect and contractive ones with negative states. However, cultural variations significantly influence ; the thumbs-up conveys positivity in Western cultures but is offensive in parts of the and , highlighting the need for context-aware models in global applications. Techniques for recognition typically involve skeleton-based processing, where 2D video analysis extracts keypoints from RGB footage, contrasting with 3D systems that use depth sensors for precise spatial reconstruction. Temporal sequences of these skeletons are modeled using recurrent neural networks (RNNs) or (LSTM) units to capture sequential dependencies in gestures, such as the progression from neutral to agitated motion. Seminal work has demonstrated that LSTM-enhanced RNNs achieve robust of like and by processing joint orientations and velocities over time. As of 2025, advances include datasets like BER2024 for training recognition systems classifying expressions into categories such as negative, neutral, pain, and positive, alongside hybrid models integrating s with other modalities for improved accuracy in affective computing. In human-computer interaction (HCI), these methods enable tracking for adaptive interfaces, such as agents responding to user via detected slumped s. Laboratory evaluations report accuracies of 60-80% for recognizing basic emotions (e.g., , ) using skeleton data, with higher rates (up to 92%) for distinct categories like in controlled settings. Challenges include ambiguity, where the same (e.g., crossed arms) may reflect comfort or depending on , and in multi-person scenarios, which obscures keypoints and reduces model reliability. fusion with other cues can mitigate these issues, though analysis remains essential for silent or occluded environments.

Data and Modeling

Datasets and Databases

Datasets and databases play a crucial role in affective computing by providing annotated resources for training and evaluating systems. These resources vary by , capturing speech, facial expressions, physiological signals, or multimodal data, and are essential for developing models that generalize across diverse emotional contexts. Key datasets often include acted or spontaneous expressions, with annotations for categorical emotions (e.g., , ) or dimensional models (e.g., valence-arousal).

Speech Datasets

Speech-based datasets focus on prosodic features like , , and to infer emotions from audio recordings. The Berlin Emotional Speech Database (Emo-DB) is a seminal acted dataset featuring ten speakers expressing seven German emotions, such as anger, boredom, and joy, through 535 utterances recorded in a controlled . A 2025 update, EmoDB 2.0, extends the original with additional naturalistic recordings for improved real-world applicability. It has been widely used for benchmarking speech emotion recognition due to its high-quality recordings and balanced emotional categories. Another influential resource is the Interactive Emotional Motion Capture database (IEMOCAP), which includes improvised and scripted dialogues from ten speakers in interactions, annotated for nine emotions including neutral, happy, and sad, with over 12 hours of data emphasizing spontaneous expressions.

Facial Datasets

Facial expression datasets provide image or video sequences labeled for , often derived from action units or direct categorical annotations. The Extended Cohn-Kanade dataset (CK+) contains 327 posed image sequences from 123 subjects, depicting seven basic (anger, contempt, disgust, fear, happiness, sadness, surprise) with peak frames annotated for action units, making it suitable for controlled expression analysis. For in-the-wild scenarios, AffectNet offers over one million facial images crowdsourced from the , labeled for eight discrete plus continuous and dimensions, with approximately 450,000 manually annotated images to support robust training on naturalistic variations in lighting, pose, and occlusion.

Physiological Datasets

Physiological signal datasets capture bio-signals like EEG, ECG, and GSR to detect emotion-induced and . The Database for Emotion Analysis using Physiological signals (DEAP) records EEG, peripheral physiological measures (e.g., GSR, ), and facial videos from 32 participants viewing 40 one-minute music videos, annotated on a 9-point - , totaling 32 experimental sessions for music-elicited emotions. The Wearable Stress and Affect Detection dataset (WESAD) features multimodal data from 15 subjects using wrist- and chest-worn sensors during baseline, (Trier Social Stress Test), amusement, and meditation conditions, with ECG, EDA, and signals labeled for three affective states to enable wearable-based detection.

Multimodal Datasets

Multimodal datasets integrate multiple channels for comprehensive modeling. The SEMAINE database comprises recordings of interactions between humans and limited agents, totaling approximately 80 hours, annotated for continuous dimensions (, , power, expectation) during emotionally colored conversations with 150 participants. The BAUM-1 dataset includes 1,184 spontaneous audio-visual video clips from 31 subjects reacting to affective video stimuli, labeled for five emotions (, , , , neutral) and mental states, capturing upper-body gestures and facial expressions in naturalistic settings.
DatasetModalityKey FeaturesSize/DetailsPrimary Source
Emo-DBSpeechActed German emotions (7 categories)535 utterances, 10 speakersBurkhardt et al. (2005)
IEMOCAPSpeech/MultimodalImprovised dyadic dialogues (9 emotions)12+ hours, 10 speakersBusso et al. (2008)
CK+FacialPosed sequences (7 emotions, action units)327 sequences, 123 subjectsLucey et al. (2010)
AffectNetFacialIn-the-wild images (8 emotions + V-A)1M+ images, ~450K annotatedMollahosseini et al. (2017)
DEAPPhysiologicalEEG/GSR/video for music-induced V-A32 participants, 40 trialsKoelstra et al. (2011)
WESADPhysiologicalWearable sensors for stress/affect15 subjects, 4 conditionsSchmidt et al. (2018)
SEMAINEMultimodalAudio-visual interactions (continuous V-A)959 recordings, 150 participantsMcKeown et al. (2012)
BAUM-1MultimodalSpontaneous AV for emotions/mental states1,184 clips, 31 subjectsZhalehpour et al. (2017)
Annotation of affective datasets presents significant challenges due to the subjective nature of , leading to variability in labeling. , often measured by Cohen's or , typically ranges from 0.4 to 0.7 in emotion annotation tasks, indicating moderate agreement influenced by cultural differences and annotator ; for instance, Kappa scores below 0.6 highlight the need for multiple annotators and protocols to ensure dataset quality. Ethical sourcing requires explicit from participants, particularly for sensitive emotional , to address risks and potential psychological distress, with guidelines emphasizing anonymization and minimization in collection protocols.

Algorithms and Computational Models

In affective computing, algorithms and computational models process emotional data to infer affective states, often integrating features from sources like speech, facial expressions, and physiological signals. Feature-level fusion, an early approach, combines raw or extracted features from multiple modalities before , such as concatenating Mel-frequency cepstral coefficients (MFCCs) from audio with action units (AUs) from facial landmarks to create a unified representation. This method enhances complementarity across modalities by merging them at the input stage, with a common technique involving weighted summation defined as f_{fused} = \sum w_i f_i, where f_i are individual modality features and weights w_i are optimized through methods like grid search to balance contributions and mitigate noise. Decision-level fusion, in contrast, performs late integration by aggregating independent predictions from each modality, such as through voting mechanisms or probabilistic models that assess modality confidence. Bayesian networks are particularly effective here, modeling dependencies between modality outputs to compute a joint posterior probability for the final emotion label, enabling robust handling of missing or unreliable data from one source. This approach assumes conditional independence among modalities post-feature extraction, often yielding higher accuracy in noisy real-world scenarios compared to unimodal systems. Machine learning techniques form the backbone of these models, with support vector machines (SVMs) widely used for binary classification due to their effectiveness in high-dimensional spaces. SVMs map features to a that maximizes the margin between positive and negative valence classes, achieving accuracies up to 70-80% on physiological datasets by selecting optimal kernels like radial basis functions. For sequential data, (LSTM) networks capture temporal dependencies in emotional expressions, trained with loss L = -\sum y \log(\hat{y}), where y is the true label and \hat{y} the predicted probability, minimizing divergence between distributions over time steps. Recent trends as of 2025 emphasize multimodal transformers, such as those augmenting fusion with cross-attention mechanisms, which outperform LSTMs by jointly modeling long-range interactions across modalities with accuracies exceeding 85% on benchmark tasks. Dimensional modeling shifts from categorical to continuous representations, using to predict (pleasantness) and () dimensions. models, expressed as v = \beta_0 + \beta_1 x, where v is the valence score and x the feature , provide a simple yet interpretable baseline, often extended to nonlinear variants like support for capturing complex affective dynamics. These models facilitate fine-grained analysis, with mean absolute errors as low as 0.15 on valence scales when trained on annotated datasets, enabling applications in . Personalization enhances model generalizability by adapting to individual differences, commonly via , where pre-trained models on large corpora are fine-tuned with user-specific data. This unsupervised domain adaptation transfers knowledge from source domains (e.g., general emotion datasets) to target users, reducing and improving accuracy by 10-20% in subject-dependent scenarios through techniques like feature alignment or adversarial training. Such methods ensure computational models remain effective across diverse populations without extensive relabeling.

Theoretical Frameworks

Affective computing draws on several psychological theories to model human computationally, providing foundational concepts for systems that recognize and respond to affective states. These frameworks emphasize the cognitive, social, and dimensional aspects of emotions, guiding the design of emotion-aware technologies without delving into implementation details. The cognitivist approach posits emotions as outcomes of cognitive appraisals of environmental stimuli, where individuals evaluate situations relative to their . In Richard S. Lazarus's 1991 model, this process involves primary appraisal, assessing whether an event threatens or benefits the individual, followed by secondary appraisal, evaluating coping resources, which together determine the emotional response. This framework has influenced affective computing by framing emotions as relational and adaptive, enabling models that simulate appraisal-based decision-making in agents. In contrast, the interactional approach views as socially constructed phenomena shaped by context and cultural influences rather than innate universals. Lisa Feldman Barrett's constructed theory, articulated in , argues that emotions emerge from the brain's predictive processing of interoceptive signals and external cues, categorized through learned concepts influenced by social and environmental factors. This perspective challenges fixed emotion categories in computational systems, advocating for flexible, context-dependent representations that account for variability across individuals and cultures. Hybrid models integrate cognitivist and interactional elements into computational structures, such as the (Fuzzy Logic Adaptive Model of Emotions) framework, which uses to map appraisals of events to emotional intensities like , , or . Developed in 2000, FLAME operationalizes appraisal dimensions—such as desirability, expectedness, and —within agent-based simulations, allowing dynamic emotion adaptation based on ongoing interactions. Theoretical frameworks also distinguish between categorical and dimensional models of emotion representation. Categorical theories assume discrete emotion types, such as or , while dimensional models represent emotions in continuous spaces; and James A. Russell's Pleasure-Arousal-Dominance (PAD) model (1974, with applications in 1996) provides a three-dimensional framework where pleasure reflects positivity, arousal indicates activation level, and dominance captures control perceptions. PAD has been widely adopted in affective computing for its ability to map diverse emotional states vectorially, facilitating valence-based analysis. Critiques of these frameworks highlight cultural biases, particularly in Western-centric models like Lazarus's appraisals or Ekman-inspired categorizations, which may overlook non-Western expressions of , leading to skewed computational interpretations. In the , research has shifted toward inclusive frameworks that incorporate data and contextual variability, promoting emotion models sensitive to diverse populations to mitigate biases in affective systems.

Applications

Healthcare and Mental Health

Affective computing plays a pivotal role in healthcare by enabling objective detection and monitoring of emotional states to support diagnostics, , and care, particularly for conditions where subjective reporting can be unreliable. In detection, vocal prosody analysis has emerged as a key application, identifying acoustic features such as reduced pitch variability that correlate with depressive symptoms. Studies show that depressed individuals often exhibit lower pitch variability, slower speech rates, and altered prosody, which models can detect with accuracies ranging from 70% to 80% in benchmarks like the Audio/Visual Emotion Challenge (AVEC). For instance, studies using vocal prosody analysis have achieved up to 79% accuracy in classifying from speech samples. For autism therapy, social robots like Kaspar facilitate training by engaging children in interactive scenarios that model basic emotions such as happiness and sadness. Long-term studies involving approximately 170 children with demonstrate that Kaspar's structured play sessions improve social interaction, collaborative behaviors, and skills, with participants showing enhanced ability to identify expressions compared to baseline assessments. These interventions leverage the robot's predictable responses to build without overwhelming sensory input. Pain assessment in non-verbal patients benefits from physiological signal monitoring, including (HRV) and galvanic skin response (GSR), which provide objective indicators of acute distress. HRV decreases and GSR increases during painful episodes, enabling wearable devices to detect with accuracies exceeding 80% in multimodal setups combining these signals. This approach is particularly valuable for patients unable to self-report, such as those in intensive care, and helps mitigate misuse by guiding precise dosing based on real-time data rather than subjective estimates, potentially reducing over-prescription risks. Mental health applications incorporate AI chatbots like Woebot, which employ alongside (CBT) principles to deliver personalized interventions. Woebot analyzes user text inputs for emotional tone, guiding sessions on mood tracking, thought challenging, and coping strategies, resulting in moderate reductions in and anxiety symptoms as measured by standardized scales like the PHQ-9. Meta-analyses indicate that remote monitoring in digital mental health interventions, including affective computing applications, can improve outcomes compared to traditional methods, with stronger effects in guided interventions for and anxiety through consistent emotional tracking and timely feedback.

Education and Personalized Learning

Affective computing has significantly transformed by enabling intelligent tutoring systems that adapt to students' emotional states, thereby enhancing and learning efficacy. One prominent example is AutoTutor, an intelligent tutoring system developed to simulate human-like conversations in , which detects learner through cues including facial expressions, , and conversational patterns. By monitoring these indicators, AutoTutor dynamically adjusts the difficulty of prompts and provides targeted interventions, such as simplifying explanations or offering encouragement, to mitigate negative affect and promote deeper understanding in subjects like Newtonian physics and . In e-learning environments, engagement monitoring via physiological signals like galvanic skin response (GSR) plays a crucial role in maintaining student attention. GSR sensors, which measure changes in skin conductance associated with emotional arousal, allow platforms to detect waning focus or disinterest during online sessions and trigger real-time motivational prompts, such as interactive questions or personalized summaries. This approach has been integrated into adaptive e-learning systems to foster sustained involvement, particularly in remote or self-paced learning scenarios, by responding to subtle physiological shifts that precede behavioral disengagement. Personalized feedback mechanisms further exemplify affective computing's impact, utilizing on student-generated text like essays to identify emotional undertones and provide supportive responses. For instance, techniques can detect indicators of anxiety, such as repetitive phrasing or negative sentiment markers in writing, enabling tutors to offer empathetic guidance or resources to alleviate without derailing academic progress. This method supports emotional alongside content mastery, as seen in applications where automated analysis flags affective patterns in student submissions to inform tailored interventions. Empirical studies from the to 2025 demonstrate the tangible benefits of these adaptive platforms, particularly those employing affective detection, which have reported learning retention boosts of 10-25% compared to traditional methods. For example, trials with emotion-sensitive review systems in MOOCs showed average improvements in learning gains of about 21.8% on weekly assessments, attributed to timely affective adaptations that enhanced knowledge retention and reduced dropout rates by up to 14.3% in distance learning contexts. These findings underscore the role of affective computing in optimizing educational outcomes across diverse settings. To promote inclusivity, affective computing extends to emotion-aware () simulations designed for neurodiverse learners, such as those with or ADHD. These systems use real-time from , physiological, and behavioral data to customize VR environments, adjusting sensory inputs or narrative pacing to match individual emotional tolerances and thereby improving and participation in immersive learning experiences. Such applications have shown promise in enhancing engagement for neurodivergent students by providing safe, adaptive spaces that simulate social or academic scenarios with immediate affective feedback.

Human-Computer Interaction and Assistive Technologies

Affective computing has significantly enhanced human-computer interaction (HCI) by enabling systems to detect and respond to users' emotional states, thereby improving and intuitiveness in everyday interfaces. Empathetic interfaces, such as desktop agents, utilize physiological signals like galvanic skin response (GSR) to identify user levels and dynamically adjust the interaction pace—for instance, slowing down animations or simplifying prompts to alleviate during tasks. This approach, pioneered in early affective computing research, allows software agents to exhibit by adapting to emotional cues, fostering a more supportive in prolonged computing sessions. In assistive technologies, emotion-detecting prosthetics integrate affective signals with traditional mechanisms to enable more intuitive for users with limb differences. For example, systems fuse myoelectric signals from residual muscles with affective data derived from bioelectric measurements, such as frontal face electrodes capturing emotional states, to modulate prosthetic responses—allowing the device to align movements with the user's , like gentler grips during calm states or firmer ones under . This co-adaptive reduces the and enhances precision, as demonstrated in evaluations where affective integration improved task performance by adapting to mental states in . Virtual reality (VR) applications leverage affective avatars to strengthen emotional connections in telepresence scenarios, where remote users interact through digital representations that convey non-verbal cues. These avatars detect and mirror users' via integrated sensors, sharing affective data in real-time to simulate natural —such as displaying subtle expressions or posture shifts that enhance social bonding during meetings. Research shows that such systems increase perceived emotional interdependence and , making telepresence more engaging for collaborative or personal interactions without physical proximity. Accessibility in HCI benefits from affective computing through tailored for deaf users, which interprets emotional nuances in gestures and facial expressions to facilitate clearer communication in digital interfaces. Tools like ASL systems process video inputs to identify affective markers, enabling software or interpreters to convey tone alongside literal signs, thus bridging gaps in . Studies indicate that these technologies reduce in HCI tasks for deaf individuals by streamlining interpretation and response times, with some implementations showing notable efficiency gains in metrics. Brief integration with further refines accuracy without dominating the process. As of 2025, emerging trends in affective computing emphasize integration with augmented reality (AR) glasses for real-time social cue augmentation, where devices overlay emotional insights—such as highlighting subtle facial expressions or suggesting empathetic responses—directly into the user's field of view. This enhances everyday social interactions by providing proactive affective support, particularly for neurodiverse users or in high-stakes conversations, drawing on generative AI to personalize cues while maintaining privacy through on-device processing. Authoritative reviews highlight this as a high-impact direction, with prototypes demonstrating improved social navigation in diverse settings.

Transportation and Automotive Systems

Affective computing plays a crucial role in and automotive systems by enabling vehicles to detect and respond to drivers' and passengers' emotional states, thereby enhancing and . These systems integrate sensors such as cameras, physiological monitors, and microphones to recognize affective signals like or , allowing for real-time interventions that mitigate risks associated with , which contributes to approximately 90% of accidents. In particular, driver monitoring systems (DMS) leverage affective computing to adapt vehicle controls, issue alerts, or adjust environmental features, promoting safer driving in both manual and autonomous contexts. One primary application is the detection of driver drowsiness and , which affective computing addresses through facial and physiological analysis. Facial recognition techniques measure metrics like PERCLOS (Percentage of Eyelid Closure Over the Pupil Over Time), where eyes are considered more than 80% closed during evaluation periods, providing a validated indicator of decline. Systems alert drivers when PERCLOS exceeds thresholds such as 70%, as demonstrated in studies evaluating for characterization, triggering vibrations, audio cues, or seat adjustments to restore . Physiological signals, briefly referencing monitored via wearables, complement these visual cues for robust detection in varied conditions. Affective computing also targets in traffic scenarios to predict and prevent behaviors. Galvanic response (GSR) and (HRV) sensors detect elevated levels indicative of , with models using these signals to forecast risky maneuvers like sudden . For instance, increased GSR conductivity and HRV alterations have been linked to emotional preceding aggressive actions, enabling predictive interventions. Upon detection, vehicles can automatically adjust in-car elements, such as playing calming music or modifying climate control to lower cabin temperature, thereby reducing driver tension and promoting smoother operation. For passengers, particularly in autonomous vehicles, affective computing enhances comfort through emotion-aware adaptations. Cabin-mounted cameras analyze expressions to infer states like or anxiety, facilitating personalized selections such as recommending videos or podcasts that match detected moods. This approach ensures a more engaging ride by dynamically curating content, improving overall satisfaction during long journeys where drivers are not actively engaged. Notable case studies illustrate the impact of these technologies. In the 2020s, Ford's Driver Alert system and Mercedes-Benz's Attention Assist employed affective monitoring to detect via steering patterns and cues, contributing to reported reductions of around 15% through timely affect-based alerts in fleet trials. Similarly, regulations under the Act, effective from 2024, classify emotion AI in vehicles as high-risk, mandating and risk assessments to ensure safe deployment while prohibiting misuse in non-safety contexts. Despite these advancements, challenges persist in affective computing for automotive applications. Privacy concerns arise in shared rides, such as ride-hailing services, where continuous of multiple occupants risks unauthorized and breaches of . Additionally, accuracy suffers in diverse lighting conditions, with facial recognition systems experiencing up to 20% error rates in low-light or glare scenarios, necessitating robust preprocessing techniques like infrared augmentation to maintain reliability.

Entertainment and Gaming

Affective computing has transformed entertainment and gaming by enabling systems to detect and respond to players' emotional states, thereby creating more immersive and personalized experiences. In adaptive narratives, adjust story elements based on inferred player emotions, often derived from physiological signals or behavioral cues. For instance, an game employs an experience manager that modifies trajectories according to detected emotions such as distress, fear, hope, or joy, enhancing emotional resonance and player investment. Similarly, procedural adapt character representations and difficulty levels using physiological like skin conductance, allowing narratives to evolve in real-time to match the player's affective response. Character AI in gaming leverages affective computing to simulate realistic emotional interactions with non-player characters (NPCs), particularly in role-playing games (RPGs). A seminal example is the 2005 game , an interactive drama where NPCs engage in real-time dialogue that responds to the player's verbal and emotional inputs, using to convey relational tensions and affective dynamics such as or . This approach draws on simulation techniques to model emotional states, enabling NPCs to exhibit believable affective behaviors that deepen narrative immersion without relying on scripted responses. In virtual reality (VR) and augmented reality (AR) environments, affective computing integrates haptic feedback to amplify emotional experiences, such as delivering jolts to simulate fear during intense moments. Haptic vests and devices provide tactile cues synchronized with emotional detection from biofeedback, like heart rate variability, to heighten immersion in fitness or adventure games. For example, in VR crowd scenarios, kinesthetic haptics enhance emotional responses by simulating physical interactions, leading to stronger feelings of anxiety or excitement compared to visual-only feedback. Biofeedback-driven adaptations in AR fitness games further tailor challenges to maintain optimal arousal levels, promoting sustained engagement through emotionally attuned progression. Studies on player engagement metrics demonstrate the impact of , with tracking via physiological sensors showing significant improvements in and retention. A 2024 methodological study on affect-adaptive action games reported a notable increase in self-reported positive and enjoyment, indicating enhanced emotional satisfaction over non-adaptive controls. Systematic reviews confirm that approximately 65% of affect-adaptive implementations yield positive effects on player experience, including higher and prolonged play sessions, as measured by tools like the Game Experience Questionnaire. These findings underscore how emotional can extend gameplay duration by aligning content with individual affective profiles. In the broader industry, affective computing influences entertainment beyond games, as seen in Disney's development of emotionally expressive . Using and motion-capture pipelines, Disney robots incorporate affective cues like dynamic gaze and posture to convey emotions such as curiosity or sadness, reducing design time from years to months while enabling lifelike interactions in theme parks. Similarly, streaming platforms like employ in recommendation systems to infer user moods from viewing patterns and , suggesting content that aligns with emotional states—such as uplifting comedies for low-mood periods—to boost personalization and retention. These applications highlight affective computing's role in scaling across media ecosystems.

Challenges and Future Directions

Technical and Methodological Challenges

One major technical challenge in affective computing is achieving high accuracy in emotion recognition due to individual differences among users. Studies have shown that classification accuracy can vary by up to 20% based on factors such as age and gender, particularly when analyzing psychobiological signals like skin conductance and heart rate variability across different emotional states. These variations arise because emotional expressions and physiological responses differ systematically; for instance, elderly males often exhibit higher recognition accuracy compared to young females in certain valence-arousal conditions. Additionally, overfitting in training datasets exacerbates accuracy issues, as models trained on limited or non-diverse data fail to generalize, leading to significant disparities in performance across demographic groups. Robustness to real-world conditions remains a critical limitation, with environmental noise severely impacting recognition performance. In speech-based emotion detection, factors like urban background noise or accents degrade acoustic features such as mel-frequency cepstral coefficients, causing substantial drops in accuracy due to masked emotional cues and the , where speakers alter vocal production in noisy settings. Similarly, for analysis, variations in conditions introduce interference that reduces model reliability. Performance often plummets when transitioning from controlled environments—where accuracies range from 65% to 97% for and other affective states—to unconstrained real-world scenarios, sometimes approaching near-zero effectiveness in detecting subtle via wearables. This gap highlights the need for models resilient to heterogeneous contexts, as lab-trained systems struggle with low-intensity responses and sensor limitations in daily use. Scalability poses another hurdle, particularly in techniques that integrate data from text, audio, and visuals for comprehensive analysis. approaches, such as transformers and CNN-LSTM hybrids, incur high computational costs, requiring significant GPU resources for and limiting deployment on devices. Fine-grained strategies, while improving representation quality, create a with efficiency, as cross-attention mechanisms demand substantial and power during and . Efforts to mitigate this include lightweight architectures like and quantization, but these often trade off some accuracy for reduced . The lack of in labeling and datasets further complicates methodological . There is no universal framework for annotating emotions, with datasets employing varied categorical models (e.g., Ekman's six basic emotions) or dimensional scales (e.g., 3- to 7-point valence-arousal ratings), resulting in inconsistencies like differing inter-annotator agreements from 0.43 to 0.91. This variability hinders comparability across studies and model training. Moreover, many datasets suffer from limited diversity in demographics, contexts, and modalities, prompting recent calls—particularly in 2025 reviews—for more inclusive, large-scale collections to capture varied emotional expressions and reduce biases. Interoperability challenges arise when integrating affective computing systems with existing human-computer frameworks, especially setups in domains like healthcare. Seamless data exchange between new models and older infrastructures is often impeded by incompatible protocols and formats, complicating deployment in clinical or assistive environments. This lack of increases costs and risks data silos, undermining the potential for emotional in practical applications.

Ethical and Societal Considerations

Affective computing systems often involve constant monitoring of users' emotional states through physiological signals, facial expressions, or voice patterns, raising significant risks due to the sensitive nature of emotional . For instance, breaches in wearable devices that collect biometric for affect recognition can expose personal emotional histories, leading to potential or . In 2023, surveys indicated that nearly 60% of consumers expressed concerns over vulnerabilities in connected devices, including fitness trackers that infer emotions from , highlighting the growing threat of unauthorized access to such . Bias in affective computing arises primarily from underrepresentation in training datasets, resulting in disparate performance across demographics. Facial expression recognition models, for example, demonstrate significantly lower accuracy—often 20-30% reduced—for non-Western or faces compared to ones, due to skewed data favoring Western populations. This representational can perpetuate inequities in applications like hiring or , where misread emotions lead to unfair outcomes. Mitigation strategies include fairness-aware techniques, which adjust models to balance performance across protected attributes like and . The potential for through affective computing poses ethical challenges, particularly in contexts where affect prediction is used to without adequate . In , systems analyze emotional responses to tailor content, potentially exploiting vulnerabilities to drive purchases, as seen in targeted campaigns that predict and respond to user or excitement in . issues are amplified in public spaces, where emotion-sensing cameras or apps may process data without explicit user awareness, undermining and raising concerns about psychological . Societally, affective computing could contribute to job displacement in roles requiring , such as counseling, by automating emotional support through companions that simulate understanding. While these tools offer scalable assistance, they risk devaluing human interaction in therapeutic settings, potentially leading to workforce shifts where handles routine empathy tasks. Additionally, over-reliance on for emotional support has been linked to diminished and increased dependency, with studies showing users forming attachments to chatbots that may exacerbate rather than foster genuine connections. Regulatory guidelines are emerging to address these concerns, with the EU AI Act of 2024 classifying systems as high-risk or prohibited in sensitive contexts like workplaces and to protect . The Act mandates transparency, risk assessments, and human oversight for such technologies to prevent misuse. Complementing this, the IEEE has developed standards for ethical considerations in emulated , emphasizing principles like beneficence, non-maleficence, and in the design and deployment of affective systems. One prominent emerging trend in affective computing involves the integration of large models (MLLMs) with advanced architectures to enable more nuanced, contextual emotion understanding. For instance, the 2025 has advanced this area by leveraging large models to shift from categorical to generative methods that incorporate data such as text, audio, and visuals for improved affect interpretation. Similarly, EmoVerse, a framework introduced in 2025, enhances MLLMs by addressing limitations in traditional , allowing for better alignment of linguistic reasoning with emotional cues in human-computer interactions. AffectGPT further exemplifies this trend, providing a new benchmark dataset and model that utilizes MLLMs to elevate beyond simplistic , achieving higher accuracy in generating emotion-aware responses. Advancements in brain-computer interfaces (BCIs), particularly those employing (EEG), are opening new avenues for direct decoding of affective states, with significant implications for . Recent developments in 2025 have demonstrated EEG-BCI systems capable of affect decoding from brain signals, enabling applications in restorative that respond to users' emotional intentions for enhanced control and feedback. For example, hybrid EEG-BCI paradigms have been explored to decode complex emotional patterns, supporting neuroprosthetic devices that adapt to affective variations in clinical settings like . These innovations build on trends toward non-invasive decoding, promising more intuitive interfaces for individuals with motor or cognitive impairments. Edge AI is gaining traction for on-device affective processing, prioritizing privacy and real-time responsiveness in resource-constrained environments. The 2025 CloudCom special session on affective computing in cloud-edge architectures highlights collaborative paradigms that enable local emotion analysis on wearables and mobile devices, minimizing data transmission to central servers and thereby enhancing user privacy. This approach supports low-latency applications, such as real-time emotion monitoring in personal assistants, by processing multimodal inputs directly on edge hardware without compromising sensitive affective data. Efforts in are addressing biases in through datasets like , which facilitate analysis of and related affective cues in diverse contexts. The dataset, comprising annotated video clips from English-language TV shows, has been instrumental in training models for detection that integrate verbal, visual, and acoustic modalities, serving as a foundation for extending affective computing to multilingual and scenarios. Recent reviews in 2025 emphasize its role in recognition, where techniques improve understanding of culturally nuanced s like irony, paving the way for inclusive systems. Systematic analyses further underscore the need for such datasets to incorporate multilingual elements, enhancing and detection across linguistic boundaries. Despite these advances, open challenges persist in long-term affect modeling, particularly in incorporating stable personality traits for sustained emotional prediction. Research from 2022 onward has highlighted the potential of continuous frameworks that track affective variations over time to infer personality influences, yet scaling these to real-world, longitudinal applications remains complex due to inter-individual differences. Benchmarks for computing, including audio-visual models, reveal gaps in robust, long-term modeling that accounts for trait-based emotional baselines, calling for integrated approaches in future affective systems. Additionally, in low-power devices poses a key hurdle, as affective computing's computational demands strain battery life and . A 2025 study on sustainable Emotion-AI proposes optimizations for sentiment analysis on edge devices, reducing environmental impact through lightweight models that maintain accuracy while operating on minimal power. These efforts aim to balance affect with eco-friendly designs, ensuring broader deployment without excessive resource consumption.

References

  1. [1]
    None
    Summary of each segment:
  2. [2]
    Affective Computing | Books Gateway - MIT Press Direct
    According to Rosalind Picard, if we want computers to be genuinely intelligent and to interact naturally with us, we must give computers the ability to ...
  3. [3]
    Overview ‹ Affective Computing - MIT Media Lab
    The Affective Computing group creates and evaluates new ways of bringing together Emotion AI and other affective technologies in order to make people's lives ...Missing: history | Show results with:history
  4. [4]
    Twenty-Five Years of Research in Affective Computing
    Sep 29, 2025 · Affective Computing is a cross-disciplinary field of research aimed at creating machines that can recognize, interpret, simulate and ...Missing: history | Show results with:history
  5. [5]
    Affective Computing - MIT Press
    This book provides the intellectual framework for affective computing. It includes background on human emotions, requirements for emotionally intelligent ...Missing: PDF | Show results with:PDF
  6. [6]
    [PDF] Affective Computing - Cornell | ARL
    This paper presents and discusses key issues in “affective computing,” computing that relates to, arises from, or influences emotions. Models are sug- gested ...
  7. [7]
    Affective Computing: Recent Advances, Challenges, and Future ...
    Jan 5, 2024 · The third aspect is sentiment analysis, which utilizes machine-learning and deep-learning algorithms to model and identify emotional signals.
  8. [8]
    Universal Emotions | What are Emotions? - Paul Ekman Group
    Ekman identified the six basic emotions as anger, surprise, disgust, enjoyment, fear, and sadness. His research shows the strongest evidence to date of a ...Universal Facial Expressions · Atlas of Emotions · What is Anger? · Books
  9. [9]
    [PDF] A Circumplex Model of Affect
    Factor-analytic evidence has led most psychologists to describe affect as a set of dimensions, such as displeasure, distress, depression, excitement, ...Missing: valence- | Show results with:valence-
  10. [10]
    Kismet - The First Social Robot - MIT Media Lab
    Kismet is the first robot designed to explicitly engage people in natural and expressive face-to-face interaction.Missing: affective | Show results with:affective
  11. [11]
    Speech emotion recognition combining acoustic features and ...
    In this paper we introduce a novel approach to the combination of acoustic features and language information for a most robust automatic recognition of a ...Missing: Bjorn | Show results with:Bjorn
  12. [12]
    A review of affective computing: From unimodal analysis to ...
    In this paper, we not only discuss the multimodal fusion of unimodal information, but also discuss the state of the art in unimodal affect recognition methods, ...Missing: 2010s | Show results with:2010s
  13. [13]
    AAAC – Association for the Advancement of Affective Computing
    The AAAC is a professional, world-wide association for researchers in Affective Computing, Emotions and Human-Machine Interaction.Missing: Humaine | Show results with:Humaine
  14. [14]
    MER 2025: When Affective Computing Meets Large Language Models
    Apr 28, 2025 · MER2025 is the third year of our MER series of challenges, aiming to bring together researchers in the affective computing community to explore emerging trends.
  15. [15]
    Emotion Recognition Using Different Sensors, Emotion Models ...
    Correct recognition of human emotions can promote the development of affective computing. Most existing emotion recognition surveys only focus on a single ...Emotion Recognition Using... · 3. Sensors For Emotion... · 5. Classification
  16. [16]
    Appraisal Models | The Oxford Handbook of Affective Computing
    Within affective computing research, models derived from appraisal theories of emotion emphasize appraisal as the central process to be modeled.
  17. [17]
    Emotions are emergent processes: they require a dynamic ...
    Scherer K. R.1984On the nature and function of emotion: a component process approach. In Approaches to emotion (eds Scherer K. R., Ekman P.), pp. 293–317 ...
  18. [18]
  19. [19]
    [PDF] a meta-analysis on unimodal and multimodal affect detection ...
    Oct 26, 2012 · Multimodal affect detection accuracies were consistently better than unimodal, with an average 8.12% improvement, and 29.4% higher than the ...
  20. [20]
    Multi-Modal Fusion Emotion Recognition Method of Speech ...
    Jul 8, 2021 · The experimental results show that the recognition accuracy of the proposed method on the MOSI and MELD datasets are 87.56 and 90.06%, ...
  21. [21]
    Performance Metrics for Multilabel Emotion Classification - MDPI
    This study compares various F1-score variants—micro, macro, and weighted—to assess their performance in evaluating text-based emotion classification.
  22. [22]
    Cultural similarities and differences in display rules
    Matsumoto, D., & Ekman, P. (1989a). American-Japanese cultural differences in judgments of facial expressions of emotion.Motivation and Emotion, 13 143–157.
  23. [23]
    Cultural Modes of Expressing Emotions Influence How ... - NIH
    Culture is known to affect ideals for emotional arousal (Tsai, 2007) and norms for emotional display (Matsumoto, 1990), as well as how often and how strongly ...
  24. [24]
    GANimation: Anatomically-aware Facial Animation from a Single ...
    GANs are a powerful class of generative models based on game theory. A typical GAN optimization consists in simultaneously training a generator network to ...Missing: affective | Show results with:affective
  25. [25]
    Haptic Empathy: Conveying Emotional Meaning through Vibrotactile ...
    May 8, 2021 · The objective of this research is to investigate whether people can share subjective feelings through simple vibrotactile feedback.
  26. [26]
    A neural approach to the Turing Test: The role of emotions
    The Turing Test proposes the possibility of distinguishing the behavior of a machine from that of a human being through an experimental session.Missing: believability variants
  27. [27]
    Perceived Empathy of Technology Scale (PETS) - ACM Digital Library
    May 11, 2024 · The PETS allows designers and researchers to evaluate and compare the perceived empathy of interactive systems rapidly.Missing: believability Turing
  28. [28]
    Robots are both anthropomorphized and dehumanized when ...
    Aug 5, 2024 · Such findings raise concerns amongst some who worry that emotion-simulating robots are psychologically manipulative. In the current study, we ...
  29. [29]
    Survey on speech emotion recognition: Features, classification ...
    Another challenging issue is that how a certain emotion is expressed generally depends on the speaker, his or her culture and environment. Most work has focused ...
  30. [30]
    Speech emotion recognition using machine learning — A systematic ...
    Traditional features include the prosodic, spectral, and acoustic features ... SVM scheme for speech emotion recognition using MFCC feature. International ...Missing: seminal | Show results with:seminal
  31. [31]
    [PDF] Acoustic Feature Analysis in Speech Emotion Primitives Estimation
    In this paper we focus on emotion recognition from speech signals. A majority of research on speech emotion recognition clas- sifies emotions into a small ...
  32. [32]
    Speech Emotion Recognition Via CNN-Transformer and ... - arXiv
    We propose a Speech Emotion Recognition network based on CNN-Transformer and multi-dimensional attention mechanisms.
  33. [33]
    Emotion Recognition from Speech Using Wav2vec 2.0 Embeddings
    This work proposes a transfer learning method for speech emotion recognition where features extracted from pre-trained wav2vec 2.0 models are modeled using ...Missing: HMM | Show results with:HMM
  34. [34]
    A review on speech emotion recognition: A survey, recent advances ...
    Aug 9, 2025 · Among these, speech emotion recognition (SER) remains central yet challenging, as it is hindered by speaker variability, contextual and cultural ...
  35. [35]
    Modelling individual and cross-cultural variation in the mapping of ...
    Jan 16, 2023 · The existence of a mapping between emotions and speech prosody is commonly assumed. We propose a Bayesian modelling framework to analyse ...
  36. [36]
    Benchmarking Pretrained Models for Speech Emotion Recognition
    The RAVDEES dataset achieved an accuracy of 77.02%, whereas the IEMOCAP and Emo-DB datasets achieved accuracies of 72.25% and 85.57%, respectively. The use ...
  37. [37]
    Multiple Acoustic Features Speech Emotion Recognition Using ...
    Experiments conducted on the IEMOCAP benchmark dataset have shown that our proposed system can achieve a 73.80% weighted accuracy (WA) and 74.25% unweighted ...
  38. [38]
    Speech emotion recognition using fine-tuned Wav2vec2.0 and ...
    We conducted speech emotion recognition experiments on the IEMOCAP dataset. The experiments show that our model achieves the weighted accuracy of 73.37% and the ...Missing: benchmarks | Show results with:benchmarks
  39. [39]
    Fusing Speech Emotion and Text Sentiment Using Machine Learning
    This study introduces a novel framework for enhancing emotional understanding by fusing speech emotion recognition (SER) and sentiment analysis (SA).
  40. [40]
    Combining speech-based and linguistic classifiers to recognize ...
    Jan 31, 2019 · Our approach merges textual sentiment analysis and emotion recognition from paralinguistic features to respectively analyze the text ...
  41. [41]
    Action unit classification for facial expression recognition using ...
    Apr 4, 2021 · In this study, we utilized active learning and support vector machine (SVM) algorithms to classify facial action units (AU) for human facial expression ...
  42. [42]
    Facial landmarks with dlib, OpenCV, and Python - PyImageSearch
    Apr 3, 2017 · The pre-trained facial landmark detector inside the dlib library is used to estimate the location of 68 (x, y)-coordinates that map to facial ...
  43. [43]
    Micro-expression recognition model based on TV-L1 optical flow ...
    Oct 20, 2022 · We pre-train the model with optical flow features to make it easier to capture the micro-movements of facial micro-expressions. We integrate ...Micro-expression recognition... · Classification of pre-processed... · Experiments
  44. [44]
    Facial Action Coding System - APA PsycNet
    Ekman, P., & Friesen, W. V. (1978). Facial Action Coding System (FACS) [Database record]. PsycTESTS. https://doi.org/10.1037/t27734-000 ...Missing: original | Show results with:original
  45. [45]
  46. [46]
    [PDF] Recognizing Facial Expression: Machine Learning and Application ...
    We present a systematic comparison of machine learning methods applied to the problem of fully automatic recogni- tion of facial expressions.
  47. [47]
    A Component Based Approach Improves Classification of Discrete ...
    Jul 18, 2010 · The aim of this study is to simulate the pareidolia capability of humans to produce an emotional response to a scene using analysis of facial ...Missing: deep | Show results with:deep
  48. [48]
    [PDF] Facial Expression Recognition with Deep Learning - CS230
    It is one of the more challenging datasets with human-level accuracy only at 65±5% and the highest performing published works achieving 75.2% test accuracy.
  49. [49]
    [PDF] Automatic Recognition of Facial Actions in Spontaneous Expressions
    Abstract— Spontaneous facial expressions differ from posed expressions in both which muscles are moved, and in the dy- namics of the movement.
  50. [50]
    Differences in Facial Expressions between Spontaneous and Posed ...
    Feb 21, 2020 · The results indicate that spontaneous smiles have higher intensities for upper face than lower face. On the other hand, posed smiles showed higher intensities ...
  51. [51]
    Advances in facial expression recognition technologies for emotion ...
    Sep 23, 2025 · Lightweight architectures such as MobileNet and EfficientNet have been explored for on-device FER in mobile and edge environments. Vision ...
  52. [52]
    A Real-Time and Privacy-Preserving Facial Expression Recognition ...
    Jul 16, 2024 · This study proposes an edge computing-based facial expression recognition system that is low cost, low power, and privacy preserving.
  53. [53]
    Wearable-Based Affect Recognition—A Review - PMC - NIH
    Therefore, the aim of this paper is to provide a broad overview and in-depth understanding of the theoretical background, methods and best practices of wearable ...
  54. [54]
    Emotion recognition systems with electrodermal activity
    We conducted a systematic review and meta-analysis on electrodermal-activity-based emotion-recognition systems. Our findings suggest that arousal prediction ...Missing: seminal | Show results with:seminal
  55. [55]
    Emotion Recognition Systems with Electrodermal Activity
    Emotion Recognition Systems with Electrodermal Activity: From Affective Science to Affective Computing ... The paper introduces a multimodal affective ...Missing: seminal | Show results with:seminal
  56. [56]
    Stress and Heart Rate Variability: A Meta-Analysis and Review of ...
    This review aimed to survey studies providing a rationale for selecting heart rate variability (HRV) as a psychological stress indicator.
  57. [57]
    Facial EMG – Investigating the Interplay of Facial Muscles ... - NCBI
    Nov 29, 2022 · Facial EMG is a method to investigate emotions and affect, uncovering subtle emotional components and revealing suppressed emotions.
  58. [58]
    Automated Emotion Recognition System Using Blood Volume Pulse ...
    Jun 29, 2023 · In this study, a new method for detecting emotions using Blood Volume Pulse (BVP) signals and machine learning was presented.
  59. [59]
    Facial color is an efficient mechanism to visually transmit emotion
    Mar 19, 2018 · Our hypothesis is that a face can send emotion information to observers by changing the blood flow or blood composition on the network of blood ...Missing: pallor | Show results with:pallor
  60. [60]
    Smartwatches in healthcare medicine: assistance and monitoring
    Nov 3, 2023 · They showed that motion artifacts highly affect the reliability of HRV parameters. In the study conducted by Sarhaddi et al. (2022) [66] ...
  61. [61]
    Emotion Recognition from Skeletal Movements - PMC - NIH
    Jun 29, 2019 · The process of using a RNN and RNN-LSTM for gestures-based emotion recognition sequence of movement is presented in Figure 6b. 3. Results and ...
  62. [62]
    CMU-Perceptual-Computing-Lab/openpose - GitHub
    OpenPose has represented the first real-time multi-person system to jointly detect human body, hand, facial, and foot keypoints (in total 135 keypoints) on ...
  63. [63]
    [PDF] Expressing emotion through posture and gesture
    Emotions are expressed through body posture, like a collapsed posture when depressed, and gestures, such as leaning forward to show interest. Posture is ...
  64. [64]
    [PDF] Survey on Emotional Body Gesture Recognition
    Lift the right or left hand up. Finger point with right or left hand. Finger or hand shaky. Arms crossing. Sadness. Body dropped. Shrunk body. Bowed shoulders.
  65. [65]
    Body language in different cultures around the world: A top guide
    Aug 21, 2023 · While the thumbs-up gesture typically means "good" or "okay" in many cultures, it's considered offensive in Iran and can be taken as a vulgar ...
  66. [66]
    Decoding spatiotemporal features of emotional body language in ...
    Sep 5, 2022 · The present data revealed that overall emotion recognition was high. Anger sequences were categorized with the highest accuracy (M = 91.9%, SEM ...Results · Emotion Recognition Of Full... · Feature Importance For...
  67. [67]
    (PDF) Emotion Recognition from Skeletal Movements - ResearchGate
    The proposed algorithm creates a sequential model of affective movement based on low level features inferred from the spacial location and the orientation of ...
  68. [68]
    Skeleton-Based Emotion Recognition Based on Two-Stream Self ...
    We propose a self-attention enhanced spatial temporal graph convolutional network for skeleton-based emotion recognition.Missing: RNN | Show results with:RNN
  69. [69]
    [PDF] A Review of Constraints on Vision-based Gesture Recognition for ...
    This paper surveys major constraints on vision-based gesture recognition occurring in detection and pre-processing, representation and feature extraction, and ...
  70. [70]
    [PDF] A Comprehensive Survey on Affective Computing - arXiv
    May 8, 2023 · Affective computing is only considered in the context of interdisciplinary fields like psychology and computer science, along with their ...
  71. [71]
    [PDF] IEMOCAP: Interactive emotional dyadic motion capture database
    Nov 9, 2008 · To facilitate such investigations, this paper describes a new corpus named the “interactive emotional dyadic motion cap- ture database” (IEMOCAP) ...
  72. [72]
    [PDF] The Extended Cohn-Kanade Dataset (CK+) - Iain Matthews
    In 2000, the Cohn-Kanade (CK) database was released for the purpose of promoting research into automatically detecting individual facial expressions.
  73. [73]
    AffectNet: A Database for Facial Expression, Valence, and Arousal ...
    Aug 14, 2017 · AffectNet is by far the largest database of facial expression, valence, and arousal in the wild enabling research in automated facial expression recognition.
  74. [74]
    [PDF] DEAP: A Database for Emotion Analysis using Physiological Signals
    Abstract—We present a multimodal dataset for the analysis of human affective states. The electroencephalogram (EEG) and peripheral physiological signals of ...<|separator|>
  75. [75]
    [PDF] Introducing WESAD, a Multimodal Dataset for Wearable Stress and ...
    In this paper we present a novel dataset for stress and affect detection. The subjects (n = 15) were exposed to different affective stimuli (stress and ...
  76. [76]
    [PDF] The SEMAINE Database: Annotated Multimodal Records of ...
    The SEMAINE database contains multimodal recordings of emotionally colored conversations between people and simulated agents, including human-human and human- ...
  77. [77]
  78. [78]
    Inter-Rater Reliability for Emotion Annotation in Human-Computer ...
    Aug 6, 2025 · In this paper we investigate the achieved inter-rater agreement utilizing Krippendorff's alpha for emotional annotated interaction corpora and ...
  79. [79]
    Ethics Sheet for Automatic Emotion Recognition and Sentiment ...
    Jun 9, 2022 · This ethics sheet fleshes out assumptions hidden in how AER is commonly framed, and in the choices often made regarding the data, method, and evaluation.
  80. [80]
    [PDF] Review on Multimodal Fusion Techniques for Human Emotion ...
    Multimodal fusion techniques for emotion recognition include feature-level, decision-level (late fusion), and hybrid approaches, using multiple modalities like ...<|separator|>
  81. [81]
    Enhanced multi-modal emotion recognition using the feature level ...
    Multi-modal human emotion recognition is a complex process of synthesizing information from various modalities to calculate emotion states.
  82. [82]
    Physio-visual data fusion for emotion recognition - ScienceDirect.com
    For decision-level fusion, we have implemented two methods; the first is based on voting process and the second is based on dynamic Bayesian Networks. The ...
  83. [83]
    Advances in Multimodal Emotion Recognition Based on Brain ...
    This paper primarily discusses the progress of research into multimodal emotion recognition based on BCI and reviews three types of multimodal affective BCI ( ...
  84. [84]
    Application of SVM-RFE on EEG signals for detecting the most ...
    ▻ SVM-RFE is proposed for feature selection related to affective valence. ▻ Topography-time–frequency analysis is used to get relevant scalp regions ...
  85. [85]
    SAE+LSTM: A New Framework for Emotion Recognition From Multi ...
    Jun 11, 2019 · To recognize emotion using the correlation of the EEG feature sequence, a deep neural network for emotion recognition based on LSTM is proposed.
  86. [86]
    Multimodal transformer augmented fusion for speech emotion ...
    May 21, 2023 · We propose a method named multimodal transformer augmented fusion that uses a hybrid fusion strategy, combing feature-level fusion and model-level fusion ...Missing: EmoBERTa | Show results with:EmoBERTa
  87. [87]
    Predicting Valence and Arousal from Affective Images
    Jun 25, 2025 · The results imply that the CNN regression model performs better when predicting emotional dimensions than the Random Forest regression model.
  88. [88]
    Predicting the Arousal and Valence Values of Emotional States ...
    Jul 7, 2024 · We exploit the Remote Collaborative and Affective Interactions (RECOLA) database to predict arousal and valence values using machine learning techniques.
  89. [89]
    [PDF] Personalizing EEG-Based Affective Models with Transfer Learning
    In this paper, we propose to personalize EEG-based affec- tive models by adopting two kinds of domain adaptation ap- proaches in an unsupervised manner. One is ...
  90. [90]
    [PDF] A Survey on Personalized Affective Computing in Human-Machine ...
    Apr 1, 2023 · This approach can help to improve the accuracy and effectiveness of machine learning systems by tailoring them to the specific characteristics ...
  91. [91]
    Emotion and Adaptation - Richard S. Lazarus - Oxford University Press
    The work provides a complete theory of emotional processes, explaining how different emotions are elicited and expressed, and how the emotional range of ...
  92. [92]
    The theory of constructed emotion: an active inference account of ...
    In this article, we begin with the structure and function of the brain, and from there deduce what the biological basis of emotions might be. The answer is a ...
  93. [93]
    FLAME—Fuzzy Logic Adaptive Model of Emotions
    In this paper, we propose a new computational model of emotions that can be incorporated into intelligent agents and other complex, interactive programs.
  94. [94]
    Pleasure-arousal-dominance: A general framework for describing ...
    Evidence bearing on the Pleasure-Arousal-Dominance (PAD) Emotional State Model was reviewed and showed that its three nearly orthogonal dimensions provided.
  95. [95]
    Domain adaptation for bias mitigation in affective computing
    Mar 17, 2025 · Biases in technologies like sentiment analysis (SA) and facial emotion recognition (FER) can result in adverse societal impacts and diminished ...
  96. [96]
    Automated Audiovisual Depression Analysis - PMC - PubMed Central
    That said, classification accuracy for detecting depression has been above chance, typically falling between 70% and 80%, with some studies reporting as high as ...
  97. [97]
    Speech Features as Predictors of Momentary Depression Severity in ...
    Jan 18, 2024 · Pitch variability, speech pauses, and speech rate were associated with depression severity, positive affect, valence, and energetic arousal.
  98. [98]
    [PDF] Detecting Depression from Facial Actions and Vocal Prosody
    Accuracy for vocal prosody was 79%. These findings suggest the feasibility of automatic detection of depression, raise new issues in automated facial image ...
  99. [99]
    A Review of Socially Assistive Robotics in Supporting Children with ...
    This study aimed to investigate the use of social robots as an interactive learning approach for treating children diagnosed with autism spectrum disorder ...
  100. [100]
    Kaspar, the social robot and ways it may help children with autism
    Aug 6, 2025 · Kaspar is a child-sized, socially-interactive humanoid robot developed by the Adaptive Systems Research Group at the University of Hertfordshire (UK)
  101. [101]
    Heart rate monitoring to detect acute pain in non-verbal patients
    Apr 14, 2023 · To evaluate HR as a tool to assist pain management for non-verbal patients, other well-established biomarkers of pain, such as heart rate ...
  102. [102]
    Electrodermal activity in pain assessment and its clinical applications
    Aug 6, 2024 · Electrodermal activity (EDA) measures skin conductivity, reflecting sweat gland activity, and is considered a noninvasive measure of the ...
  103. [103]
    Exploration of physiological sensors, features, and machine learning ...
    Jul 9, 2021 · This subjective pain assessment results in suboptimal treatment plans, over-prescription of opioids, and drug-seeking behavior among patients.
  104. [104]
    [PDF] Woebot Health Research Bibliography
    Automated conversational agents can deliver CBT content through text-based conversations, reducing depression and anxiety symptoms in select populations. This ...
  105. [105]
    Delivering Cognitive Behavior Therapy to Young Adults With ...
    Jun 6, 2017 · A fully automated conversational agent to deliver a self-help program for college students who self-identify as having symptoms of anxiety and depression.Cited by (1530) · Authors · Tweetations (296) · MetricsMissing: sentiment | Show results with:sentiment
  106. [106]
    Effect of Engagement With Digital Interventions on Mental Health ...
    Nov 3, 2021 · This is the first review providing empirical evidence that engagement with DMHIs is associated with therapeutic gains.<|separator|>
  107. [107]
    The evolving field of digital mental health: current evidence and ...
    May 15, 2025 · The above‐mentioned recent meta‐analysis showed that effects were larger in trials that delivered apps based on CBT principles (compared to ...
  108. [108]
    AutoTutor Detects and Responds to Learners Affective and ...
    Jun 23, 2008 · The affect-sensitive AutoTutor detects the emotions (boredom, flow/engagement, confusion, frustration) of a learner by monitoring conversational ...
  109. [109]
    AutoTutor and affective autotutor: Learning by talking with ...
    Multimodal semi-automated affect detection from conversational cues, gross body language, and facial features. ... Facial features for affective state detection ...
  110. [110]
    AutoTutor detects and responds to learners affective and cognitive ...
    The new versions of AutoTutor detect learners' boredom, confusion, and frustration by monitoring conversational cues, gross body language, and facial features.<|control11|><|separator|>
  111. [111]
    [PDF] Measuring Emotion in Education Using GSR and HR Data from ...
    This paper embarks on an exhaustive exploration of the intricate emotional tapestry woven by students across diverse academic levels within the confines of face ...Missing: seminal | Show results with:seminal
  112. [112]
    Affective Computing for Learning in Education: A Systematic Review ...
    The galvanic skin response (GSR) is an interesting channel endemic to emotion research. Changes in sweat gland activity are measured to indicate the ...
  113. [113]
    Real-Time monitoring of ECG and GSR signals during computer ...
    The findings of this experiment influence the appropriateness of instructional intervention, and drive the development of real-time assessment for education.
  114. [114]
    [PDF] Measuring Student Emotions in an Online Learning Environment
    The motivation of this work is to demonstrate that from an approach using sentiment analysis it is possible to know the affective aspects of the students, ...
  115. [115]
    Application of Affective Computing in Sentiment Analysis of English ...
    The results showed that there is no significant difference between students' English writing level and their emotional tendency in the text, but compared ...
  116. [116]
    Affective computing in education: A systematic review and future ...
    This study presents a review of the literature on affective computing in education by selecting articles published from 2010 to 2017.Missing: essays | Show results with:essays
  117. [117]
    Adaptive Review for Mobile MOOC Learning via Multimodal ...
    In a 3-week study involving 28 learners, we found that AttentiveReview2 on average improved learning gains by 21.8% in weekly tests. Follow-up analysis shows ...
  118. [118]
    Design of an integrated multi-modal machine learning framework for ...
    Aug 27, 2025 · Adaptive learning pathways produced by the RL-ALP component enhance distance learning outcomes by 14.3% and decrease dropout rates by ...Missing: percentage | Show results with:percentage
  119. [119]
    Adaptive Neuro-Affective Engagement via Bayesian Feedback ...
    This paper presents an adaptive neuro-affective system designed to enhance engagement in children with neurodevelopmental disorders through serious games.
  120. [120]
    Adaptive Neuro-Affective Engagement via Bayesian Feedback ...
    Oct 10, 2025 · This paper presents an adaptive neuro-affective system designed to enhance engagement in children with neurodevelopmental disorders through serious games.
  121. [121]
    Affective Computing In Virtual Reality - Meegle
    Education and Training: In educational settings, VR simulations can adapt to students' emotional responses, making learning more engaging and effective.
  122. [122]
    Affective Computing - MIT Press
    This book provides the intellectual framework for affective computing. It includes background on human emotions, requirements for emotionally intelligent ...
  123. [123]
    [PDF] Studying Stress Using Skin Conductance - Hal-Inria
    This paper reports an experiment for stress recognition in human- computer interaction. Thirty-one healthy participants performed five stressful. HCI tasks and ...Missing: desktop agents example
  124. [124]
    Co-Adaptive and Affective Human-Machine Interface for Improving ...
    Bioelectric signals are recorded using one pair of electrodes placed on the frontal face region of a user to extract the mental (affective) measures (the ...
  125. [125]
    [PDF] Remote Sharing Non-Verbal Affective Data through 2D Animated ...
    Abstract—This paper describes the use of an affective avatar that detects and shares non-verbal affective data in a 2D-based remote meeting application.
  126. [126]
    MULTICOLLAB-ASL: Towards Affective Computing for the Deaf ...
    Oct 27, 2024 · This manuscript explores preliminary findings from an ongoing multimodal ASL corpus collection and analysis study, focusing on human-generated ...
  127. [127]
    [PDF] Human-Centered Editable Speech-to-Sign-Language Generation ...
    Jun 24, 2025 · In studies with 20 deaf signers and 5 professional interpreters, we observe a +13 point SUS improvement, 6.7 point reduction in cognitive load, ...
  128. [128]
    [PDF] Narrative Review of Emotional Expression Support in XR - arXiv
    May 1, 2025 · In ex- ploring the use of smart glasses equipped with an emotion recognition system to enhance human-to-human communication, with a focus on ...<|control11|><|separator|>
  129. [129]
    Empathic Extended Reality in the Era of Generative AI
    A key enabler of XR as an empathic entity is the integration of affective computing, which allows XR systems to recognize users' emotional and cognitive states ...
  130. [130]
    Driver Emotion Recognition for Intelligent Vehicles: A Survey
    Jun 17, 2020 · We overview each study's methodology to measure and recognize emotions in the context of driving. Across the literature, we find a strong ...
  131. [131]
    Affectiva Automotive AI for Driver Monitoring Systems
    Affectiva Automotive AI takes driver state monitoring to the next level, analyzing both face and voice for levels of driver impairment.
  132. [132]
    PERCLOS-based technologies for detecting drowsiness
    The percentage of time that the eyes are more than 80% closed (PERCLOS) is one of the most validated indices used for the passive detection of drowsiness, which ...
  133. [133]
    Driver Drowsiness Detection Based on Face Feature and PERCLOS
    To characterize driver's fatigue based on eyelid movement, they used the percentage of eye closure during a certain time interval (PERCLOS). They verified that ...
  134. [134]
    A Review on Measuring Affect with Practical Sensors to Monitor ...
    Using sensors to monitor signals produced by drivers is a way to help better understand how emotions contribute to unsafe driving habits.
  135. [135]
    Real-World Driver Stress Recognition and Diagnosis Based ... - NIH
    This paper proposes different CNN and CNN-LSTSM-based fusion models using physiological signals (SRAD dataset) and multimodal data (AffectiveROAD dataset)
  136. [136]
    [PDF] A Machine Learning Approach to Predict Emotional Arousal and ...
    This study aims to predict aggressive driving behavior and emotional arousal using physiological signals,. Electrodermal Activity (EDA) and Heart Rate (HR), ...
  137. [137]
    Emotion-aware Design in Automobiles: Embracing Technology ...
    Apr 25, 2025 · Thus, their systems aim to adjust in-car settings dynamically (e.g., climate control, adaptive music, ambient light, or empathic speech) to ...<|control11|><|separator|>
  138. [138]
    Persona-PhysioSync AV: Personalized Interaction through ...
    Mar 20, 2024 · A significant aspect of the PPS-AV framework is its real-time monitoring of passenger engagement and comfort levels within AVs. It considers a ...
  139. [139]
    Emotion Detection AI Creating the Ultimate Self-Driving Car ...
    Feb 2, 2024 · Emotion detection AI comes in – it will be key to transforming self-driving vehicles into friendly, caring companions that passengers truly enjoy riding in.
  140. [140]
    How do I use the Driver Alert System in my Ford?
    The Driver Alert System can monitor your level of alertness based on your driving behavior. If the system detects that your driving pattern has become ...
  141. [141]
  142. [142]
    Regulating Emotional Artificial Intelligence in Cars (Chapter 15)
    Dec 7, 2024 · The chapter also argued how emotional AI in cars is likely to be regulated in the new EU regulations. ... European Parliament in April 2024 ...
  143. [143]
    Affective Computing Research - Meegle
    One of the most significant challenges in affective computing is ensuring user privacy. Since these systems rely on sensitive data such as facial ...
  144. [144]
    Advancing driver fatigue detection in diverse lighting conditions for ...
    Jul 10, 2024 · This study introduces a novel method for driver fatigue detection aimed at enhancing the safety and reliability of intelligent driving assistance systems.
  145. [145]
    Theories, methodologies, and effects of affect-adaptive games
    Rooted in the discipline of affective computing, originally described by Picard [11], affect-adaptive games try to directly elicit emotional responses by ...
  146. [146]
    Emotion in Games | The Oxford Handbook of Affective Computing
    People choose to play games as a “voluntary attempt to overcome unnecessary obstacles” (Suits, 2005), as play is among the main motivators for learning, mental ...
  147. [147]
    [PDF] Façade: An Experiment in Building a Fully-Realized Interactive Drama
    Although their visual fidelity continues to in- crease, computer-controlled non-player characters (NPCs) remain stubbornly simplistic, with a narrow repertoire ...
  148. [148]
    Haptic feedback as affective amplifier: enhanced fear perception ...
    Jul 23, 2025 · Haptic feedback serves as a potent affective amplifier in virtual reality (VR), intensifying threat perception to influence emotional intensity.
  149. [149]
    Haptic feedback in a virtual crowd scenario improves the emotional ...
    Nov 27, 2023 · This study aims to investigate the impact of kinesthetic haptics on eliciting emotional responses within crowded virtual reality (VR) scenarios.
  150. [150]
    Advancing Methodological Approaches in Affect-Adaptive Video ...
    Oct 28, 2024 · Here we present a methodological approach to explicitly test mapping assumptions between in-game variables, player input, and subjective experience to build ...
  151. [151]
    Adaptation in Affective Video Games: A Literature Review
    Aug 7, 2025 · Video games with affect-based adaptation measure player's behavior signals and recognize player's emotional states in order to adapt specific game features.
  152. [152]
    How Disney Packed Big Emotion Into a Little Robot - IEEE Spectrum
    Oct 6, 2023 · Disney's robotic character uses a new programming pipeline to bring emotional movements into the real world.
  153. [153]
    A Survey of Affective Recommender Systems: Modeling Attitudes ...
    Aug 27, 2025 · Affective Recommender Systems are an emerging class of intelligent systems that aim to enhance personalization by aligning recommendations with ...
  154. [154]
    How Netflix uses emotional analytics to improve CX - TechTarget
    Dec 5, 2019 · Netflix uses emotional analytics to craft a better customer experience by efficiently shortening movie trailers for consumers to evaluate movies more quickly.Missing: affective computing
  155. [155]
    Affective Computing and the Impact of Gender and Age | PLOS One
    In summary, the conducted classification processes resulted in 20% classification accuracy differences according to age and gender, especially when comparing ...Missing: variance | Show results with:variance
  156. [156]
    Affective Computing and Emotional Data - arXiv
    Sep 25, 2025 · This paper examines the technological underpinnings, regulatory considerations, and ethical implications of emotionally intelligent AI systems.
  157. [157]
    Speech Emotion Recognition in Noisy Real-World Environments
    Jun 4, 2025 · This paper critically examines the transition of SER systems from controlled laboratory settings to real-world scenarios characterized by ...
  158. [158]
    [PDF] Affect Detection From Wearables in the “Real” Wild
    Feb 13, 2023 · Affect detection from wearables in the “real” wild—where people go about their daily routines in heterogeneous contexts—is a different ...
  159. [159]
    A Comprehensive Review of Multimodal Emotion Recognition - NIH
    This paper presents a comprehensive review of multimodal emotion recognition (MER), a process that integrates multiple data modalities such as speech, ...
  160. [160]
  161. [161]
    A review and critical analysis of multimodal datasets for emotional AI
    Aug 13, 2025 · The overall Fleiss kappa scores of 0.43 and 0.91 were reported for emotion and sentiment annotation, respectively. Another version of MELD was ...
  162. [162]
    Systematic Review of Image‐Based Emotion Recognition Datasets ...
    Oct 20, 2025 · These datasets are vital for training emotion recognition models and enhancing accuracy by incorporating body language and gestures alongside ...
  163. [163]
    Request Rejected
    **Summary:**
  164. [164]
    Data privacy and security worries are on the rise, while trust is down
    Sep 5, 2023 · Nearly six in 10 respondents to our survey worry that their devices are vulnerable to security breaches (for example, hackers stealing personal data).
  165. [165]
    Fitness Tracker Privacy Risks - Kaspersky
    Wearable companies could be vulnerable to data breaches – a famous example being 2018's Under Armour MyFitnessPal breach, which exposed the usernames, passwords ...Missing: affective computing emotion 2023
  166. [166]
    The Price of Emotion: Privacy, Manipulation, and Bias in Emotional AI
    Sep 9, 2024 · This article examines the data privacy, manipulation, and bias risks of Emotional AI, analyzes relevant United States (“US”) and European Union (“EU”) legal ...Missing: computing hacks
  167. [167]
    [PDF] Investigating Bias and Fairness in Facial Expression Recognition
    This study investigates bias in facial expression recognition, where datasets lack fair demographic representation, leading to lower performance for minority ...Missing: Western | Show results with:Western
  168. [168]
    Faces of Fairness: Examining Bias in Facial Expression Recognition ...
    Feb 16, 2025 · This study investigates bias sources in FER datasets and models. Four common FER datasets—AffectNet, ExpW, Fer2013, and RAF-DB—are analyzed. The ...
  169. [169]
    Fairness-aware machine learning engineering: how far are we? - NIH
    Under this point of view, an AI solution is considered fair if its predictions do not causally depend on any protected attributes. Given all the nuances ...<|separator|>
  170. [170]
    Affective Computing in Marketing: Practical Implications and ...
    Jan 4, 2022 · Research should explore if consumers consent to their emotions being tracked by affective computing systems during interactions with the firm.
  171. [171]
    Technology Gets Emotional | Emerging Issues - BSR
    The use of affective technologies in both public and private spaces could significantly impact human dignity and autonomy, alongside other rights such privacy ...Missing: prediction | Show results with:prediction
  172. [172]
    Displaced counselors: An infinite supply of AI empathy - 壹心理yisum
    May 22, 2023 · Rosalind Picard, a leading researcher in the field of affective computing at MIT, has studied the effects of AI-simulated empathy on anger and ...
  173. [173]
    Spending Too Much Time With AI Could Worsen Social Skills
    Oct 7, 2024 · Similarly, our growing reliance on AI chatbots and agents could create emotional dependence with psychological consequences. The optimal ...
  174. [174]
    Emotional risks of AI companions demand attention - Nature
    Jul 22, 2025 · The integration of AI into mental health and wellness domains has outpaced regulation and research.
  175. [175]
    [PDF] Affective Computing and Emotional Data - arXiv
    Sep 24, 2025 · Under the AI Act, emotion recognition systems are addressed explicitly in Article 50(3), which states: “Deployers of an emotion recognition ...
  176. [176]
    Ethical Considerations in Emotion Recognition Research - MDPI
    The paper provides practical ethical recommendations for developing affective computing systems that advance the field while maintaining responsible deployment ...3.2. Privacy And Data... · 3.4. Psychological Impact On... · 3.6. Transparency And...
  177. [177]
    IEEE Standard for Ethical Considerations in Emulated Empathy in ...
    Jun 28, 2024 · This standard defines a model for ethical considerations and practices in the design, creation, and use of empathic technology, incorporating ...
  178. [178]
    EmoVerse: Enhancing Multimodal Large Language Models for ...
    Affective computing is a critical area of artificial intelligence, underpinning applications such as human–computer interaction and mental health monitoring.Missing: EmoBERTa | Show results with:EmoBERTa
  179. [179]
    AffectGPT: A New Dataset, Model, and Benchmark for Emotion...
    May 1, 2025 · Abstract: The emergence of multimodal large language models (MLLMs) advances multimodal emotion recognition (MER) to the next level—from naive ...<|separator|>
  180. [180]
    Application and future directions of brain-computer interfaces in ...
    Sep 14, 2025 · Various strategies have emerged to effectively decode these complex brain signals, particularly using EEG and ECoG. Motor intention decoding ...
  181. [181]
    Recent applications of EEG-based brain-computer-interface in the ...
    Mar 24, 2025 · This paper aims to offer a comprehensive review of recent electroencephalogram (EEG)-based BCI applications in the medical field across 8 critical areas.
  182. [182]
    (PDF) Challenges and Trends in Brain-Computer Interface Technology
    By 2025, Brain-Computer Interface (BCI) technology has become a ... Many BCI studies have focused on decoding EEG signals associated with whole ...
  183. [183]
    Affective Computing in Cloud-Edge Environments - CloudCom 2025
    This special session aims to focus on the critical direction of cloud-edge-device collaborative architecture, exploring novel paradigms urgently needed.
  184. [184]
    [PDF] THE 2025 EDGE AI TECHNOLOGY REPORT | Ceva's IP
    By processing data locally on wearables, imaging devices, and ambient sensors, edge. AI delivers real-time insights without compromising patient privacy or.<|separator|>
  185. [185]
    soujanyaporia/MUStARD: Multimodal Sarcasm Detection Dataset
    We release the MUStARD dataset, a multimodal video corpus for research in automated sarcasm discovery. The dataset is compiled from popular TV shows including ...Missing: emotion affective
  186. [186]
    A systematic review of sarcasm recognition — multimodal fusion ...
    Sep 4, 2025 · This systematic review is the first to focus on speech-based sarcasm recognition, charting the evolution from unimodal to multimodal approaches.
  187. [187]
    Continuous Emotion Recognition for Long-Term Behavior Modeling ...
    May 12, 2022 · We introduce a novel approach that gradually maps and learns the personality of a human, by conceiving and tracking the individual's emotional variations ...
  188. [188]
    An Open-Source Benchmark of Deep Learning Models for Audio ...
    Feb 8, 2024 · This paper presents the first reproducible audio-visual benchmark to provide a fair and consistent evaluation of eight existing personality ...Missing: challenges affect
  189. [189]
    Sustainable Emotion-AI: Reducing the Environmental Cost of Cross ...
    Oct 5, 2025 · Emotion AI-also known as affective computing-has become a cornerstone of modern human-machine interaction, powering applications that detect ...