Showing papers on "Spectrogram published in 1993"

A computational model of echo processing and acoustic imaging in frequency‐ modulated echolocating bats: The spectrogram correlation and transformation receiver

Abstract: The spectrogram correlation and transformation (SCAT) model of the sonar receiver in the big brown bat (Eptesicus fuscus) consists of a cochlear component for encoding the bat's frequency modulated (FM) sonar transmissions and multiple FM echoes in a spectrogram format, followed by two parallel pathways for processing temporal and spectral information in sonar echoes to reconstruct the absolute range and fine range structure of multiple targets from echo spectrograms. The outputs of computations taking place along these parallel pathways converge to be displayed along a computed image dimension of echo delay or target range. The resulting image depicts the location of various reflecting sources in different targets along the range axis. This series of transforms is equivalent to simultaneous, parallel forward and inverse transforms on sonar echoes, yielding the impulse responses of targets by deconvolution of the spectrograms. The performance of the model accurately reproduces the images perceived by Eptesicus in a variety of behavioral experiments on two-glint resolution in range, echo phase sensitivity, amplitude-latency trading of range estimates, dissociation of time- and frequency-domain image components, and ranging accuracy in noise.

Phonocardiogram signal analysis: techniques and performance comparison

Abstract: This paper presents the applications of the spectrogram, Wigner distribution and wavelet transform analysis methods to the phonocardiogram (PCG) signals. A comparison between these three methods has shown the resolution differences between them. It is found that the spectrogram short-time Fourier transform (STFT), cannot detect the four components of the first sound of the PCG signal. Also, the two components of the second sound are inaccurately detected. The Wigner distribution can provide time-frequency characteristics of the PCG signal, but with insufficient diagnostic information: the four components of the first sound, SI, are not accurately detected and the two components of the second sound, S2, seem to be one component. It is found that the wavelet transform is capable of detecting the two components, the aortic valve component A2 and pulmonary value component P2, of the second sound S2 of a normal PCG signal. These components are not detectable using the spectrogram or the Wigner distribution. Ho...

Cardiac Doppler blood-flow signal analysis. Part 2. Time/frequency representation based on autoregressive modelling.

Abstract: Doppler spectrograms obtained by using autoregressive (AR) modelling based on the Yule-Walker equations were investigated. A complex AR model using the in-phase and the quadrature components of the Doppler signal was used to provide blood-flow directions. The effect of model orders on the spectrogram estimation was studied using cardiac Doppler blood flow signals taken from 20 patients. The 'final prediction error' (FPE) and the 'Akaike's information criterion' (AIC) provided almost identical results in model-order selection. An index, the spectral envelope area (SEA), was used to evaluate the effect of window duration and sampling frequency on AR Doppler spectrogram estimation. The statistical analysis revealed that the SEA obtained from AR modelling was not sensitive to window duration and sampling frequency. This result verified the consistency of the AR Doppler spectrogram. The white-noise characteristics of the AR modelling error signal indicated that the Doppler blood-flow signal can be adequately modelled as a complex AR process. With appropriate model orders, AR modelling provided better Doppler spectrogram estimates than the periodogram.

Method and apparatus for generating sliding tapered windows and sliding window transforms

Abstract: A method and apparatus for the computation of sliding windows and sliding window Fourier transforms and spectrograms requiring fewer operations in comparison with the current art. The sliding windows are realized using infinite impulse response filters. If the impulse response of one filter is equal to the impulse response of a second filter after a period of time, the impulse responses can be canceled thereby resulting in a composite filter with an impulse response of finite length. In certain realizations, the infinite impulse filters have identical components. These components can be combined into a single component in the composite filter thereby reducing the required number of computational operations. Sliding windows thus realized are employed in the generation of sliding window Fourier transforms and spectrograms. The windows can be modulated or unmodulated. For unmodulated windows, the signal to be processed is first multiplied by a discrete oscillator tuned to the desired frequency of the sliding window Fourier transform or spectrogram. Modulated windows are realized with infinite impulse filters containing high frequency components and therefore do not require use of oscillators when used in sliding window Fourier transform and spectrogram circuitry. Sliding window Fourier transform and spectrogram circuitry, connected in parallel for evaluation of a number of frequency lines, can also share common circuitry, thereby significantly reducing computational and architectural overhead. Methods and apparatus for evaluation of temporally decimated sliding window Fourier transforms and spectrogram is achieved through iterative accumulation of intermediate results and downsampling.

Application of time-frequency analysis to machinery condition assessment

Abstract: Machinery condition has traditionally been assessed by analysis of the spectral energy density of the machine's vibration signal. Examination of time-frequency representations (TFRs) of constant-speed machinery data reveals vibration features that demonstrate variation in frequency over a short time period and thus cannot be adequately characterized by the power spectrum. These features may be used as supplemental information to assess the condition of the machine. Although the spectrogram provides a general indication of the time-varying spectrum, new representations such as the "cone-kernel" TFR reveal the existence of fine structural details inherent to the signal. >

Representations of ECG--late potentials in the time frequency plane.

Abstract: It is shown that time frequency signal analysis could be a useful technique in detecting late potentials (LP). In particular, the spectrogram, the Wigner distribution and the wavelet transform have been applied to ECG signals from patients with sustained ventricular tachycardia. Comparisons of the three algorithms reveals that much better time localization is achieved by the wavelet transform. Moreover, it is concluded that a separation between healthy subjects and patients could be obtained based upon time frequency analysis.

A method for filtering bioacoustic transients by spectrogram image convolution

Abstract: A method is presented for detecting bioacoustic transients. The desired transient-an animal call-is modeled as a sequence of frequency sweeps. Sweeps are detected by convolving a spectrogram of the signal with a kernel designed for the call of interest; convolution output is high when the call of interest is present and low other times. The method is tested on a set of bowhead whale (Balaena mysticetus) calls recorded in a noisy Arctic environment. The method detects bowhead calls well, performing better than a matched filter and a hidden Markov model on the task. Strengths and weaknesses of the method are discussed. >

Simultaneous, Continuous, On-Line and Real-Time Spectral Analysis of Multiple Physiologic Signals by a Personal-Computer-Based Algorithm

Abstract: We presented in this communication a computer algorithm that offers the advantage of simultaneous analysis of the power spectrum of multiple physiologic signals, including systemic arterial pressure,

Optimally Karhunen-Loeve-like STFT expansion of nonstationary processes

Abstract: To any nonstationary process, characterized by its correlation function, there corresponds an optimal window such that the short-time Fourier transform (STFT) coefficients are minimally correlated. The author makes this statement precise and derives a compact window optimization criterion in terms of the ambiguity function of the window and the expected ambiguity function of the process. As an application, he proposes a two-step procedure for time-varying spectral estimation: (I) window optimization based on an estimate of the expected ambiguity function, and (II) a spectrogram using the optimized window. An iterative algorithm is presented for the discrete-time window optimization, and experimental results are shown. >

Compression in the Time Domain.

Abstract: High input levels can cause saturation and "fast pumping" effects in hearing aids. The resulting distortion products and changes in speech amplitude envelope are examined in the time domain. Spectrograms reveal changes in speech elements not apparent with time domain methods.

Generalized time-frequency representation of ultrasonic signals

Abstract: Time-frequency representation of ultrasonic signals plays an important role in describing the scattering and dispersive effects in materials. Cohen's class of generalized time-frequency representation (GTFR) has been examined for ultrasonic applications. Two special cases of GTFR, Wigner-Ville distribution (WVD) and Choi-William distribution (CWD) are discussed. Due to the bilinear structure, all GTFRs' generate cross-terms for the multicomponent signals. The presence of cross-terms in the WVD of a multicomponent signal obscures the auto-terms. The cross-terms in CWD can be controlled by choosing a proper scale factor which does not effect the marginals. It is shown that the estimation of local time and frequency of ultrasonic echoes corrupted with noise can be accurately performed using CWD. Short-time Fourier transform (STFT) or spectrogram is also discussed as a class of GTFR. STFT or spectrogram does not satisfy the marginals. Simulated results show that CWD outperforms WVD and STFT because it is not marred by severe interference due to cross-terms, still able to satisfy the marginals, and it provides the best resolution in time-frequency plane

Space-Frequency Shape Inference and Segmentation of 3D Surfaces

Abstract: Image texture is useful for segmentation and for computing surface orientations of uniformly textured objects. If texture is ignored, it can cause failure for stereo and gray-scale segmentation algorithms. In the past, mathematical representations of image texture have been applied to only specific texture problems, and no consideration has been given to the models’ generality across different computer vision tasks and different image phenomena. We advocate. the space/frequency representation, which shows the local spatial frequency content of every point in the image. From several different methods of computing the representation, we pick the spectrogram. The spectrogram elucidates many disparate image phenomena including texture boundaries, texture in perspective, aliasing, zoom, and blur. Many past shape-from-texture algorithms require potentially unreliable feature detection and “magic numbers” (arbitrary parameters), and none of them were developed in the context of a more general texture-understanding system. Toward this end, we show that the spatial frequency shifts caused by perspective can be approximated by an &ne transformation which is a function of the textured surface’s surface normal. We use this relationship in three different shape-from-texture algorithms. TWO of them require no feature-finding and work on the raw spectrogram, giving a high-level scene parameter directly from low-level image data. The first algorithm includes an analytical sensitivity analysis. The third algorithm works with just the peak frequencies and gives a fast way of computing local surface normals from periodic texture. The algorithms need only a few magic numbers. On real textures, the average error in computed surface normal is only about four degrees. We use the third algorithm to solve a long-standing problem in image texture analysis: segmenting images of textured, 3D surfaces. Past work in texture segmentation and shape-from-texture is based on assumptions that make this problem impossible to solve. 3D perspective effects warp the otherwise uniform textures so segmentation fails. We develop a region-growing algorithm that accounts for the surface normals by unwarping the frequency distribution. It uses a minimum description length merge criterion. Our algorithm successfully segments images of real texture. We conclude by showing how the space/frequency representation has the potential for unifying several different computer vision algorithms

Application of autoregressive spectral analysis for ultrasound attenuation: interest in highly attenuating medium

Abstract: A second-order autoregressive (AR2) model, whose parameters are estimated with the Burg algorithm, is used to estimate the center-frequency along echo signals and its evolution versus depth. Data simulation of independent A-lines reflected by a homogeneous medium of scatterers are generated by a computer model with attenuation values ranging from 1 to 5 dB/cm MHz and an ultrasonic frequency of 5 MHz. The performance of the estimator is evaluated for duration of time windows ranging from 5.12 to 0.32 ms and different spectral sampling. The comparison is made with the classical Fourier spectrogram technique (FFT). It is found that the AR model provides a better estimation of attenuation than the Fourier technique: the relative error of attenuation is below 5% for windows between 0.64 to 2.56 ms, while the one obtained with the Fourier technique lies between 10 and 70% for the same window sizes. These results offer promises for determining attenuation in biological medium which are highly attenuating either because of their structure, like bone, or because high frequencies are used

Efficient STFT approximation using a quantization and differencing method

Abstract: A class of algorithms for efficiently computing approximations to the short-time Fourier transform (STFT) of any given signal is introduced. These algorithms may be classified in accordance with the number of quantization levels they use to represent the values in each short-time frame of the input signal. The structure and properties of these algorithms are illustrated through a specific case which uses a three-level quantization in each frame. Results obtained by applying an implementation of the three-level algorithm to musical and speech signals are also presented. The resulting approximations conform to the theoretical expectations of a 9-dB SNR in the STFT approximations. However, these approximations are found to require an order-of-magnitude-less computation than fast Fourier transform (FFT)-based algorithms for the exact STFT. >

Spectrogram processing of laser vibration data

Abstract: Laser vibration sensing has traditionally relied on the use of limiters and frequency modulation (FM) discriminators to process frequency modulated laser radar returns. The performance of the traditional FM discriminator approach can be limited by laser radar target characteristics and motion (speckle noise) and laser temporal coherence. In this paper we examine a novel laser vibration signal processor, a spectrogram processor, and compare its performance with the traditional limiter/FM discriminator signal processor used to process laser radar vibration measurements. The two processes are also compared using some laser radar measurement data.

Spectral features of plosives in connected-speech signals

Abstract: It is increasingly argued that continuous speech recognition systems must integrate both knowledge-based and stochastic components to optimize performance. Spectrograms are a source of acoustic-phonetic information which can be interpreted with reference to an awareness of articulatory possibilities and some of this knowledge may be formalized in terms of rules. Spectrogram interpreters appear to use an internal hierarchy of cues which can be restructured and used as part of an integrated complex of cues for identification purposes. The Speech Knowledge Interface (SKI) is used to allow a domain expert to formalize such knowledge in a machine-robust and implementable manner. This paper reports experiments and results arising from the interactive use of SKI in the identification of the place of articulation of plosives.

Deconvolution for positive time-frequency distributions

Abstract: A method is presented for generating high-resolution time-frequency distributions (TFDs) which are everywhere nonnegative and satisfy the time and frequency marginals. These TFDs are obtained by deconvolution, wherein the effects of the smoothing window are removed from a spectrogram of the signal. The method may also be used to incorporate information from multiple spectrograms of the signal, thereby improving the result. >

Analysis and filtering using the optimally smoothed Wigner distribution

Abstract: The authors consider the analysis and filtering of a deterministic signal with slowly time-varying spectra using the optimally smoothed Wigner distribution (OSWD). They compare this mixed time-frequency representation (MTFR) to other MTFRs such as the spectrogram, the short-time Fourier transform (STFT), and the Wigner and pseudo-Wigner distributions. The authors propose an approach to designing linear time-varying filters for slowly time-varying signals which is based on the concept of local nonstationarity cancellation and show that it is equivalent to masking the optimal STFT. The performance of the filter in suppressing white noise and in decomposing a slowly time-varying signal into its components is studied and compared to the performance of the techniques based on the STFT. >

High-resolution signal synthesis for time-frequency distributions

Abstract: Bilinear time-frequency distributions (TFDs) offer improved resolution over linear time-frequency representations (TFRs), but many TFDs are costly to evaluate and are not associated with signal synthesis algorithms. The spectrogram (SP) decomposition and weighted reversal correlator decomposition have been used to define low-cost, high-resolution TFDs. The authors show that the vector-valued square-root of a TFD (VVTFR) provides a representational underpinning for the TFD. By synthesizing signals from modified VVTFRs, they define high-resolution signal synthesis algorithms associated with TFDs. The signal analysis and synthesis packages can be implemented as weighted sums of SP/short-time Fourier transform (STFT) signal analysis and synthesis packages. The algorithms exhibit desirable properties, and yield results superior to STFT signal synthesis for a simple example. >

Bioacoustic transient detection by image convolution

Abstract: A method is presented for detecting bioacoustic transient signals. The method detects a common feature of animal calls, frequency sweeps, or combinations of frequency sweeps over time, by convolving a spectrogram image with a two‐dimensional kernel designed to find lines in images. This kernel has positive (excitatory) regions to achieve strong convolution operator response on desired frequency sweeps, and negative (inhibitory) regions to reject noise and interfering sounds. Design parameters control permissible variation in the sweeps. This technique is useful in screening long recordings for desired call types and perhaps for identifying individuals. For screening, it is compared with two other approaches on recordings of Bowhead Whale calls in a noisy Arctic environment: a hidden Markov model and a matched filter. This method performs better than both others on the Bowhead data, perhaps well enough for practical use. It also detects perfectly 15‐ to 20‐Hz calls of unknown origin, possibly blue whales, ...

A comparative study of spectral estimation techniques for noisy non-stationary signals with application to EEG data

Abstract: The paper considers the problem of spectral estimation of noisy non-stationary signals with application to electroencephalogram (EEG) data. Four well known methods for estimating the time-varying spectrum of a non-stationary signal are first reviewed and their performance compared. These methods which work well when the signal-to-noise ratio (SNR) is high, are shown to fail with varying degrees as SNR decreases. A technique for preprocessing noisy EEG data called time-frequency peak filtering (TFPF) is then presented and used to process EEG signals whose spectral content are highly non-stationary and difficult to model. It is shown that marked improvement in spectral estimates result after using the TFPF method. >

Time frequency representations for the analysis of speech signals

Abstract: We investigate the time-frequency representations (TFRs) for the analysis of speech signals. In order to compare the performance of some typical TFRs, e.g., spectrogram (SG), the generalized time-frequency representation with cone-shaped kernal (CK), and reduced interference distribution (RID), several experiments have been conducted using simulated signals as well as speech signals. We also formulate an algorithm for efficient computation of CK as well as RID, and present experimental results which demonstrate the potential power of the technique for speech. >

Minimum detectable signal for spectrogram displays

Abstract: Automatic detection criteria are not generally appropriate for displays where decisions are made by a human operator. However, it is shown that the Lawson-Uhlenbeck (1965) deflection criterion continues to be a useful predictor of the minimum detectable signal (MDS) for displays and simplifies the relationship between automatic and display detection performance. >

Comparison of Higher Order Moment Spectrum Estimation Techniques

Abstract: : This thesis compares the detection performance of the 1-1/2 D instantaneous power spectrum (1-1/2 D sub ips), the bispectrum, the instantaneous higher-order moment slice (IHOMS) method, and the spectrogram for multi-component stationary signals, harmonically related stationary signals, and multi-component linear FM signals corrupted by additive white Gaussian noise. In addition, a determination of the relative processing gain between the 1-1/2 D sub ips method and the spectrogram is made for stationary signals in noise. The results of this thesis show that 1-1/2 Dips has a processing gain advantage over that of the spectrogram for a range of input SNR that depends upon the size of the data window. Under some conditions, the bispectrum can detect both harmonic coupling and phase coupling between the components of multi-component signals. IHOMS' ability to detect linear chirps in noise is limited to chirps having different slew rates, and the method has a significantly greater computational cost than both the spectrogram and 1- 1/2 Dips. Bispectrum, 1-1/2 D, IPS, Higher-order moments, Cumulants

Space/frequency shape inference and segmentation of textured three-dimensional surfaces

Abstract: Image texture is useful for segmentation and for computing surface orientations of uniformly textured objects. If texture is ignored, it can cause failure for stereo and gray-scale segmentation algorithms. In the past, mathematical representations of image texture have been applied to only specific texture problems, and no consideration has been given to the models' generality across different computer vision tasks and different image phenomena. We advocate the space/frequency representation, which shows the local spatial frequency content of every point in the image. From several different methods of computing the representation, we pick the spectrogram. The spectrogram elucidates many disparate image phenomena including texture boundaries, texture in perspective, aliasing, zoom, and blur. Many past shape-from-texture algorithms require potentially unreliable feature detection and "magic numbers" (arbitrary parameters), and none of them were developed in the context of a more general texture-understanding system. Toward this end, we show that the spatial frequency shifts caused by perspective can be approximated by an affine transformation which is a function of the textured surface's surface normal. We use this relationship in three different shape-from-texture algorithms. Two of them require no feature-finding and work on the raw spectrogram, giving a high-level scene parameter directly from low-level image data. The first algorithm includes an analytical sensitivity analysis. The third algorithm works with just the peak frequencies and gives a fast way of computing local surface normals from periodic texture. The algorithms need only a few magic numbers. On real textures, the average error in computed surface normal is only about four degrees. We use the third algorithm to solve a long-standing problem in image texture analysis: segmenting images of textured, 3D surfaces. Past work in texture segmentation and shape-from-texture is based on assumptions that make this problem impossible to solve. 3D perspective effects warp the otherwise uniform textures so segmentation fails. We develop a region-growing algorithm that accounts for the surface normals by unwarping the frequency distribution. It uses a minimum description length merge criterion. Our algorithm successfully segments images of real texture. We conclude by showing how the space/frequency representation has the potential for unifying several different computer vision algorithms.

Regulating fine grinding machine - using polarisation interferometer and comparing desired calibration and actual spectrograms to create regulating variable.

Abstract: In a calibration procedure at least one spectrum of a milled material of a predetermined average grain size and/or a predetermined layer thickness or a determined dispersion grade is undertaken, for obtaining a desired spectrogram. The unit for carrying out the method includes an interferometer as an optoelectronic sensor, an amplifier (25), an ADC (26), a digital processing system (28) and a regulating stage (30) for a mill. During the subsequent milling, the corresp. measurement for obtaining a first spectrogram is carried out. At least one desired spectrogram obtained from the calibrating operation is compared with the first spectrogram for forming at least one regulating variable. This is supplied at least indirectly to the mill as an adjustment variable. ADVANTAGE - Gives increased degree of measurement accuracy.