Fact-checked by Grok 2 weeks ago

Chirp Z-transform

The Chirp Z-transform (CZT) is a computational that evaluates the of a finite sequence of length N at M points lying on a contour in the z-plane, starting from an arbitrary initial point z0 and with arbitrary angular spacing. This generalization of the (DFT) allows evaluation at unequally spaced frequencies on or off the unit circle, enabling flexible without the constraints of uniform sampling along the unit circle required by the standard DFT. The CZT achieves this efficiency by reformulating the transform as a between the input sequence and a complex exponential "chirp" signal, which is computed using fast convolution techniques like the FFT, resulting in a of approximately O((N + M) log₂(N + M)) operations for large N and M. Conceived by Leo I. Bluestein in 1968 as a linear filtering method to compute the DFT via , the algorithm was initially motivated by the need to efficiently handle prime-length or non-power-of-two DFTs using existing . It was formally developed and applied in a 1969 paper by Lawrence R. Rabiner, Ronald W. Schafer, and Charles M. Rader, who named it the "chirp Z-transform" due to the of the modulating signals w*n²/2, analogous to frequency-modulated waveforms in . This approach not only extends the DFT but also supports contours inside or outside the unit circle, making it suitable for applications requiring high-resolution frequency zooms or non-equispaced sampling. Key advantages of the CZT include its ability to compute transforms where M ≠ N, providing interpolation in the frequency domain, and its adaptability to specialized hardware or software implementations for real-time processing. Notable applications encompass high-resolution spectral estimation in signal processing, such as zoom FFTs for narrowband analysis; efficient computation of non-standard DFT sizes in digital filtering and array processing. More recent implementations as of 2025 leverage the CZT for adaptive resolution in real-time systems, including variable-bandwidth Fourier analysis.

Introduction

Definition and Purpose

The Chirp Z-transform (CZT) is a computational technique for evaluating the Z-transform of a finite-length sequence at a specified set of points along a spiral or logarithmic contour in the complex Z-plane. For an input sequence x of length N, the CZT computes the output X for k = 0, 1, \dots, M-1, where M is the desired number of output points. This contour is defined by a starting point A (a complex number representing the initial position) and a ratio W (another complex number that determines the angular and radial progression along the spiral). Mathematically, it is expressed as X = \sum_{n=0}^{N-1} x \, A^{-n} \, W^{-n k}, \quad k = 0, 1, \dots, M-1. This formulation evaluates the Z-transform X(z) = \sum_{n=0}^{N-1} x z^{-n} at the points z_k = A W^k. The primary purpose of the CZT is to enable flexible and efficient evaluation of the Z-transform at arbitrary points, including unequally spaced locations on the unit circle or other contours, without requiring a full N-point discrete Fourier transform (DFT). When A = 1 and W = e^{-j 2\pi / N}, the CZT reduces to the standard DFT, sampling uniformly along the unit circle. However, its generality allows for "zooming" into specific frequency regions or analyzing signals along non-unit contours, which is particularly useful for high-resolution spectral analysis in scenarios where uniform sampling of the DFT is insufficient or inefficient. This addresses key limitations of the DFT by permitting unequally spaced or focused points, thereby providing enhanced in targeted bands while maintaining computational through convolution-based algorithms. Such capabilities make the CZT valuable in applications requiring precise pole enhancement or in the Z-plane.

Historical Context

The Chirp Z-transform (CZT) was conceived in the fall of 1968 through a serendipitous collaboration between Lawrence R. Rabiner and Charles M. Rader at an IEEE conference in . Rader, from , suggested generalizing Leo I. Bluestein's technique using chirp signals to efficiently compute discrete transforms (DFTs) for sequences of lengths not equal to powers of two, motivated by challenges in linear filtering for . Bluestein, working at Sylvania Applied Research Laboratory, had developed this approach independently and presented it in a brief paper at the Northeast Electronics Research and Engineering Meeting (NEREM) in November 1968, titled "A Linear Filtering Approach to the Computation of the ." Rabiner shared Rader's insight with Ronald W. Schafer at Bell Laboratories, who refined and generalized the algorithm for broader applications, such as high-resolution evaluation over arbitrary contours. Their seminal work was published in as "The Chirp Z-Transform Algorithm and Its Application" in the Technical Journal, establishing the CZT as a versatile tool for non-uniform frequency sampling. The name "Chirp Z-transform" derives from the quadratic phase exponent in the algorithm, which corresponds to a discrete chirp signal—a linear frequency-modulated —used to recast the computation as a , rather than for decomposing signals into chirp components. Early adoption occurred in fields requiring flexible frequency resolution, including for estimation at , as well as and systems where non-uniform improved detection and resolution in the late and .

Mathematical Formulation

The Z-Transform

The Z-transform is a mathematical tool used in and to analyze discrete-time signals and systems, serving as the discrete-time counterpart to the continuous-time . It was introduced in the context of sampled-data systems analysis during the early 1950s. The bilateral Z-transform of a discrete-time signal x is defined as X(z) = \sum_{n=-\infty}^{\infty} x z^{-n}, where z is a in the z-plane, and the summation represents a expansion. This form applies to general two-sided sequences that may extend infinitely in both positive and negative time directions. For causal sequences, where x = 0 for n < 0, the unilateral Z-transform is used, defined as X(z) = \sum_{n=0}^{\infty} x z^{-n}. This unilateral version facilitates the analysis of systems with initial conditions and is analogous to the unilateral Laplace transform for initial-value problems. The region of convergence (ROC) for the Z-transform is the annular region in the complex z-plane where the infinite sum converges absolutely, determined by the signal's growth rate. For a right-sided causal signal like x = a^n u, the ROC is |z| > |a|, excluding possible poles at the origin. For left-sided anti-causal signals, the ROC is an interior disk |z| < 1/|a|. The ROC plays a crucial role in uniquely specifying the transform, as different signals can share the same X(z) but have distinct ROCs. If the ROC includes the unit circle |z| = 1, the inverse transform yields the discrete-time Fourier transform via evaluation on that contour. Key properties of the Z-transform enable efficient manipulation of signals and systems. Linearity states that \mathcal{Z}\{\alpha x_1 + \beta x_2\} = \alpha X_1(z) + \beta X_2(z), allowing superposition. The time-shift property gives \mathcal{Z}\{x[n - n_0]\} = z^{-n_0} X(z) for delays (n_0 \geq 0), with adjustments for advances. The convolution theorem asserts that \mathcal{Z}\{x_1 * x_2\} = X_1(z) X_2(z), linking time-domain convolution to multiplication in the z-domain, which is essential for linear time-invariant system analysis. These properties hold for both bilateral and unilateral forms, with the unilateral version incorporating initial conditions in shifts and convolutions. The Z-transform relates to other transforms by mapping the continuous-time s-plane to the z-plane via z = e^{sT}, where T is the sampling period, establishing it as the discrete analog of the for sampled systems. On the unit circle |z| = 1, it specializes to the , providing frequency-domain insights when convergence holds. The represents a sampled version of this evaluation at equally spaced points on the unit circle. For finite-length sequences, where x = 0 outside a finite interval, the Z-transform is a finite polynomial sum, converging everywhere except possibly at z = 0 or infinity, and computed directly as partial sums over the non-zero terms.

Definition of the Chirp Z-Transform

The chirp Z-transform (CZT) is a generalization of the that evaluates the transform of a finite-length discrete-time sequence at points along a specified spiral contour in the complex z-plane. For a sequence x of length N, the CZT produces output values X for k = 0, 1, \dots, M-1, where M is the desired number of output points (which may differ from N). The mathematical definition is given by X = \sum_{n=0}^{N-1} x \, A^{-n} \, W^{nk}, \quad k = 0, 1, \dots, M-1, where A is a complex scalar specifying the starting point of the contour (initial modulus and angle), and W is another complex scalar defining the geometric ratio (spiral factor) between successive points on the contour. This formulation corresponds to evaluating the unilateral Z-transform X(z) = \sum_{n=0}^{N-1} x z^{-n} at the M points z_k = A W^{-k} along the contour, which begins at z = A and spirals according to the magnitude |W| (inward if |W| < 1, outward if |W| > 1) and angular increment \arg(W). For example, in frequency-domain applications, W = e^{-j 2\pi f_0 / M} can define a spiral starting at angle \arg(A) with specified angular spacing. A special case occurs when A = 1, W = e^{-j 2\pi / N}, and M = N, in which the CZT reduces to the standard discrete Fourier transform (DFT). This derivation stems from restricting the general Z-transform to a finite sum and selecting evaluation points on the specified contour.

Computational Algorithm

Bluestein's Algorithm

Bluestein's algorithm, also known as the chirp z-transform algorithm, provides an efficient method to compute the chirp z-transform (CZT) by reformulating it as a linear convolution that can be evaluated using the fast Fourier transform (FFT). The core idea relies on a mathematical identity that rewrites the bilinear phase term nk in the CZT sum as \frac{n^2 + k^2 - (n - k)^2}{2}, transforming the quadratic phase exponentiation into a form amenable to convolution. This substitution, originally developed for computing the discrete Fourier transform (DFT) via chirp filtering, was extended to the general CZT along arbitrary contours in the complex plane. The algorithm proceeds in three main steps. First, the input sequence x for n = 0, 1, \dots, N-1 is multiplied by an input signal h = A^{-n} W^{-n^2 / 2}, yielding y = x h. Second, y is convolved with a conjugate signal g = W^{n^2 / 2}, computed efficiently via the FFT by both sequences to a L \geq N + M - 1 to prevent wrap-around effects and taking the inverse FFT of their frequency-domain product. Third, the result of the is multiplied by an output chirp signal W^{-k^2 / 2} for k = 0, 1, \dots, M-1 to obtain the CZT values X. Chirp signals are complex exponentials characterized by a quadratic , such as h = \exp\left(-j \pi n^2 / N\right) in the unit circle case where A = 1 and W = \exp(-j 2\pi / N) (assuming M = N), with L selected separately as the FFT length \geq N + M - 1. These signals introduce the necessary phase adjustments to align the CZT computation with properties. When the input length N differs from the output length M, the length L is selected as the smallest (or FFT-compatible length) at least N + M - 1 to accommodate the full linear convolution without . The original formulation appeared in 1968 for efficient DFT computation on non-power-of-two lengths using , particularly useful for zooming in radar signal processing. This was generalized in 1969 to the full CZT, enabling evaluation along spiral or circular contours beyond the unit circle.

Complexity and Efficiency

The Chirp Z-transform (CZT) algorithm achieves a of O((N + M) \log (N + M)), where N is the length of the input sequence and M is the number of output points on the contour, primarily due to the use of two fast Fourier transforms (FFTs) of length L \approx N + M - 1. This is asymptotically dominated by the FFT operations, with additional costs from pointwise multiplications by precomputed signals. In practice, the overall computation scales as L \log_2 L for moderately large N and M, making it significantly more efficient than the direct evaluation of the sum, which requires O(NM) operations. The of the CZT is O(N + M), accounting for of the input sequence, output buffers, and the signals, which can be optimized using to require approximately L/2 + 1 points for the quadratic phase factors. This linear demand supports efficient implementation without excessive overhead, particularly when is used to reach the next for FFT acceleration. A key efficiency gain of the CZT arises in scenarios requiring high-resolution spectral evaluation, where M \gg N; it avoids the prohibitive O(NM) cost of direct by leveraging FFT-based in O(N \log N) time when M is comparable to N. However, trade-offs include overhead from generating and multiplying the chirp signals, as well as zero-padding to length L, which can make the algorithm less advantageous if M is only modestly larger than N compared to standard FFT methods. Numerical stability in the CZT depends on floating-point during phase computations, particularly for the exponents in the signals w^{+n^2/2} and w^{-n^2/2}, where large n values can amplify errors, especially for contours deviating from the unit . Modern implementations mitigate this through careful scaling and higher-precision arithmetic, ensuring accuracy for a broad parameter space on the unit .

Applications and Implementations

Signal Processing Uses

The Chirp Z-transform (CZT) is widely utilized in high-resolution spectral analysis, enabling a zoom FFT that provides fine frequency resolution around signal peaks without necessitating an increase in the overall DFT size. This approach decouples spectral resolution from the length of the time record, allowing focused computation on narrow frequency bands of interest, such as 0-1.4 Hz for analyzing flight test data sampled at 50 Hz. By evaluating the z-transform along a spiral contour on the unit circle, the CZT achieves up to 10 times finer resolution than standard FFT methods while incorporating cubic interpolation to correct phase errors inherent in FFT approximations. Such capabilities are essential for measuring signal characteristics like transition slopes, bandwidth, and resolution in applications ranging from spectrum measurements to synthetic aperture radar imaging. For non-uniform frequency sampling, the CZT excels in tasks like and on arbitrary frequency grids, computing values at unequally spaced points along the unit circle with O(N log N) efficiency via Bluestein's . The interlaced CZT variant extends this by performing multiple staggered transforms and combining results to yield denser sampling in targeted regions, offering computational savings over standard zoom CZT for large datasets. This non-uniform capability supports precise in scenarios where uniform DFT sampling is inefficient, such as ultrasonic blood flow estimation or RF . In radar and sonar systems, the CZT computes spectra at unequally spaced Doppler frequencies, facilitating range cell migration correction and enhanced target in low-resolution setups. For multi-receiver synthetic aperture , it processes signals divided into range-frequency subbands, applying piece-wise linear approximations before CZT-based correction to produce high-resolution images from field data at 28 kHz frequencies. In beam forming, varying chirp rates as a function of range enable decoding of acoustic returns into beams at non-uniform frequencies, improving near-field with arrays. Audio processing benefits from the CZT in pitch detection and , where it supports logarithmic frequency spacing to align with perceptual scales, enhancing detail without uniform grid constraints. In speech enhancement, pitch-based techniques employ time-warping functions derived from CZT principles to minimize harmonic smearing, optimizing estimates and improving automatic accuracy on noisy datasets like Aurora 2. Recent advancements as of 2025 include modified CZT for extracting frequency (ENF) signals in audio tampering detection under low SNR conditions. Biomedical signals, including ECG and , leverage the CZT for focused frequency band analysis, such as refining detection by zooming into 0.8-2.2 Hz ranges corresponding to heart rates of 48-132 . This enables precise interval thresholding after bandpass filtering (5-20 Hz), achieving detection sensitivities over 99.8% on MIT-BIH databases. In analysis for biometric identification, the CZT extracts features from S1 and components via targeted spectral evaluation, supporting classification for cardiac sound authentication. A representative example is detecting signals in noise, as in remote heart rate estimation from photoplethysmography, where adaptive CZT zooms into 0.66-3 Hz bands to suppress noise and achieve mean absolute errors below 2 on datasets like UBFC-rPPG, outperforming FFT in short windows. In , zero-padded CZT enables high-accuracy parameter estimation from patterns as of May 2025.

Software and Numerical Tools

The Chirp Z-transform (CZT) is implemented in through the czt function in the Signal Processing Toolbox, which computes the length-M CZT of an input sequence x along a spiral defined by starting point A and ratio W, supporting arbitrary M including cases where M exceeds the input length N. This function integrates seamlessly with 's FFT routines for efficient computation via Bluestein's and includes built-in parameter validation to ensure valid specifications. In , provides the scipy.signal.czt function, introduced in version 1.8.0, which mirrors the API by accepting input arrays compatible with and returning the CZT along user-defined spirals, with options for axis specification and handling of M > N through zero-padding or . The scipy.signal.CZT class offers a callable for reusable transforms, optimizing repeated computations by precalculating constants. Other libraries support CZT via extensions or custom implementations based on Bluestein's convolution with FFT backends. In Julia, the FourierTools.jl package includes a Chirp Z-Transform module that leverages .jl for efficient evaluation on arbitrary contours. For C++, the pychirpz library provides a standalone implementation that can be compiled and used independently, focusing on high-performance evaluation along unit-circle arcs or spirals. itself lacks a native CZT but serves as the core for many custom C++ and Julia implementations through Bluestein-based wrappers. Open-source repositories offer accessible CZT code for educational and research purposes. The garrettj403/CZT project on GitHub implements a flexible Python-based CZT, emphasizing Bluestein's algorithm for non-uniform frequency sampling and providing examples for integration with NumPy. Key considerations in these tools include robust parameter validation for complex A and W to avoid invalid spirals, efficient handling of M > N via padded convolutions, and seamless integration with underlying FFT libraries to maintain O(N log N) complexity. Post-2020 developments include GPU-accelerated versions, such as CuPy's cupyx.scipy.signal.czt, which extends SciPy's to CUDA-enabled GPUs for large-scale computations on arrays exceeding CPU memory limits, enabling faster processing for high-resolution signals without data transfer overhead.

Relations to Other Transforms

Connection to

The Chirp Z-transform (CZT) generalizes the (DFT) by evaluating the along arbitrary contours in the complex z-plane, rather than restricting to the unit circle. Specifically, the DFT emerges as a special case of the CZT when the starting point A = 1, the angular increment W = e^{-j 2\pi / N}, and the number of output points M = N, where N is the length of the input . In this configuration, the CZT computes X = \sum_{n=0}^{N-1} x e^{-j 2\pi n k / N} for k = 0, 1, \dots, N-1, which is precisely the standard DFT formula. This inclusion highlights the CZT's broader capability: while the DFT uniformly samples the unit circle at N equally spaced points, the CZT permits flexible specification of the initial point A ( and ), the geometric ratio W (determining angular and radial steps), and an arbitrary number M of evaluation points along a spiral or . Such flexibility enables zoomed-in or non-uniform sampling without altering the core computational framework. When the contour deviates from the unit circle—achieved by setting |A| \neq 1 or |W| \neq 1—the CZT evaluates the at points inside or outside the unit , capturing effects like or that the DFT, confined to |z| = 1, cannot directly address. Historically, the CZT was developed to enable efficient computation of DFT-like transforms for non-power-of-2 lengths or focused spectral regions, building on the fast Fourier transform's success in spectrum .

Comparisons with Fast Fourier Transform

The Chirp Z-transform (CZT) offers greater flexibility than the (FFT) by evaluating the along arbitrary contours in the , such as spirals or arcs on the unit circle, allowing for non-uniform spacing and output lengths M that differ from the input length N. In contrast, the FFT computes the (DFT) on a uniform N-point grid equally spaced around the unit circle, with optimal efficiency typically requiring N to be a . This enables the CZT to focus on specific regions or achieve higher resolution in targeted bands without computing the full spectrum, whereas the FFT provides a fixed, broad from 0 to the . Both algorithms achieve O(N log N) computational complexity, but the CZT incurs additional overhead from pre- and post-multiplications by chirp signals, effectively requiring two forward FFTs, one inverse FFT, and O(N) multiplications when implemented via Bluestein's algorithm. CZT typically incurs higher runtime than the FFT due to these additional operations, though the difference varies with implementation and hardware. The FFT remains faster for full-spectrum computations due to its streamlined structure and better cache locality from fewer auxiliary operations. The FFT is preferred for broad, uniform in applications like general signal spectral estimation, where computing the entire spectrum is necessary. The CZT excels in scenarios requiring zoomed-in high-resolution analysis, such as narrowband spectrum zooming or logarithmic frequency scales, exemplified by its use in chirp-based Mel-frequency cepstral coefficients (MFCC) for audio processing to better approximate human auditory perception. Limitations of the CZT include reduced efficiency from the extra chirp multiplications and convolutions, which disrupt sequential patterns optimized in FFT libraries, making the FFT dominant for standard DFT tasks on large datasets. The CZT remains viable for large input sizes on modern hardware, supported by libraries like SciPy's signal.czt that leverage FFT overlap-add optimizations for .

References

  1. [1]
  2. [2]
    [PDF] BSTJ 48: 5. May-June 1969: The Chirp z-Transform Algorithm and ...
    To compute the z-transform along this more general contour would seem to require NM multiplications and additions since the special.
  3. [3]
    CZT vs FFT: Flexibility vs Speed - UNT Digital Library
    Oct 30, 2025 · Bluestein's Fast Fourier Transform (FFT), commonly called the Chirp-Z Transform (CZT), is a little-known algorithm that offers engineers a high- ...Missing: original | Show results with:original
  4. [4]
    "A real-time variable resolution Chirp z-Transform" by Christopher ...
    An algorithm is introduced which performs real time, high-resolution Fourier transforms of large amounts of input data over an adjustable output band.
  5. [5]
    None
    ### Summary of Chirp Z-Transform (CZT) Historical Context
  6. [6]
    Oral-History:Lawrence Rabiner
    We came up with a really ingenious theoretical procedure, which turned into the “chirp-z transform." It has been used in radar, and in sonar, and in molecular ...
  7. [7]
    [PDF] The z Transform - UTK-EECS
    May 10, 2004 · The bilateral z transform can be found using the unilateral z transform by these four steps. 1. Find the unilateral z transform and its ROC.
  8. [8]
    [PDF] Z - Transforms
    This is the Z transform, the discrete analog of the Laplace transform. As with the FT, some versions of the ZT may be one-sided n = (0, ∞) and some.
  9. [9]
    czt - Chirp Z-transform - MATLAB - MathWorks
    This MATLAB function returns the length-m chirp Z-transform (CZT) of x along the spiral contour on the z-plane defined by w and a through z = a*w.^-(0:m-1).Missing: general mathematical AW
  10. [10]
  11. [11]
    CZT — SciPy v1.16.2 Manual
    As this CZT is implemented using the Bluestein algorithm, it can compute large prime-length Fourier transforms in O(N log N) time, rather than the O(N**2) time ...<|control11|><|separator|>
  12. [12]
    Numerical error analysis of the ICZT algorithm for chirp contours on ...
    Mar 17, 2020 · This paper shows that the inverse chirp z-transform (ICZT), which generalizes the inverse fast Fourier transform (IFFT) off the unit circle ...
  13. [13]
    [PDF] High Accuracy Evaluation of the Finite Fourier Transform Using ...
    The chirp z-transform decouples the frequency resolution from the length of the time record and can place all calculated frequency points within the frequency.
  14. [14]
    [PDF] Chirp Z-Transform Spectral Zoom Optimization with MATLAB
    In using the chirp z-transform for a spectral zoom, the W and A parameters that define the z- plane contour are constrained to be scalars on the unit circle ...
  15. [15]
    The interlaced chirp Z transform - ScienceDirect.com
    In this paper, we introduce the interlaced CZT to obtain unequally spaced frequency samples. It is based on the ZCZT which computes a series of CZT's over ...Missing: original | Show results with:original
  16. [16]
  17. [17]
    A Novel Chirp-Z Transform Algorithm for Multi-Receiver Synthetic ...
    Sep 3, 2024 · We use the chirp-z transform (CZT) to correct range cell migration (RCM) to obtain low-resolution results for different subbands.Missing: early | Show results with:early
  18. [18]
    US4280203A - Sonar beam forming utilizing the chirp Z-transform ...
    The chirp rate in a chirp Z-transform sonar beam former is varied as a function of range to form beams in the near field. The chirp rate may be varied with ...
  19. [19]
    An auditory spectral analysis model using the chirp z-transform
    This paper presents a novel auditory processing model leveraging the chirp z-transform (CZT) to enhance spectral analysis in auditory signal processing.
  20. [20]
    [PDF] A Pitch-Based Spectral Enhancement Technique for Robust Speech ...
    Jun 4, 2013 · "The chirp z-transform algorithm." Audio and Electroacoustics, IEEE Transactions on 17, no. 2 (1969): 86-92. [28] Ozaktas, Haldun M., Billur ...
  21. [21]
    None
    ### Summary of Chirp Z-Transform Use in Two-Lead QRS Detection
  22. [22]
    Biometric Identification Method for Heart Sound Based on ... - MDPI
    Their method needs to locate and describe S1 and S2, then chirp-z transform (CZT) is performed to obtain the feature set, and finally, Euclidean distance ...
  23. [23]
    None
    ### Summary of Adaptive Spectral Zoom with Chirp Z-Transform for Heart Rate Detection in Noisy Biomedical Signals
  24. [24]
  25. [25]
    bionanoimaging/FourierTools.jl: Tools for working with Fourier space.
    Chirp Z-Transform; Fractional Fourier Transform; reexports FFTW.jl. Have a look in the examples folder for interactive examples. The documentation offers a ...Missing: libraries C++
  26. [26]
    thejonaslab/pychirpz: Python and C++ implementation of the Chirp-Z ...
    The Chirp-Z transform lets you evaluate any evenly-spaced set of frequencies along the unit circle (or even along an arc inside the unit circle, but we'll ...Missing: libraries Julia
  27. [27]
    FFTW Home Page
    FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions, of arbitrary input size, and of both real and ...Download Page · FFTW 3.3.10 · Upgrading from FFTW version 2 · FAQMissing: Chirp Z- Julia
  28. [28]
    garrettj403/CZT: Chirp Z-Transform - GitHub
    The chirp Z-transform (CZT) is a generalization of the discrete Fourier transform (DFT). While the DFT samples the Z plane at uniformly-spaced points along the ...
  29. [29]
    cupyx.scipy.signal.czt — CuPy 14.0.0a1 documentation
    If the transform needs to be repeated, use CZT to construct a specialized transform function which can be reused without recomputing constants. An example ...
  30. [30]
    [PDF] comparison of computing efficiency among fft, czt - arXiv
    Aug 9, 2021 · 2.3 Chirp-Z Transform. The Chirp-Z Transform (CZT) is a generalization of the DFT [15]. It is defined as follows: X[k] = N−1. ∑ n=0 x[n]A ...
  31. [31]
    [PDF] CZT vs FFT: Flexibility vs Speed Abstract - OSTI.GOV
    Bluestein's Fast Fourier Transform (FFT), commonly called the Chirp-Z Transform. (CZT), is a little-known algorithm that offers engineers a high-resolution FFT ...Missing: original paper
  32. [32]
    [PDF] The Chirp z-Transform Algorithm and Its Application
    The algorithm is based on the fact that the values of the z-transform on a circular or spiral contour can be expressed as a discrete convolution. Thus.
  33. [33]
    Significance of chirp MFCC as a feature in speech and audio ...
    A novel feature, based on the chirp z-transform, that offers an improved representation of the underlying true spectrum is proposed.