Showing papers on "Cepstrum published in 1980"

Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences

Abstract: Several parametric representations of the acoustic signal were compared with regard to word recognition performance in a syllable-oriented continuous speech recognition system. The vocabulary included many phonetically similar monosyllabic words, therefore the emphasis was on the ability to retain phonetically significant acoustic information in the face of syntactic and duration variations. For each parameter set (based on a mel-frequency cepstrum, a linear frequency cepstrum, a linear prediction cepstrum, a linear prediction spectrum, or a set of reflection coefficients), word templates were generated using an efficient dynamic warping method, and test data were time registered with the templates. A set of ten mel-frequency cepstrum coefficients computed every 6.4 ms resulted in the best performance, namely 96.5 percent and 95.0 percent recognition with each of two speakers. The superior performance of the mel-frequency cepstrum coefficients may be attributed to the fact that they better represent the perceptually relevant aspects of the short-term speech spectrum.

The theory, design, and applications of surface acoustic wave Fourier-transform processors

Abstract: This paper addresses the theory, design, and applications of surface acoustic wave (SAW) Fourier-transform processors. These analog processors are shown to perform several sophisticated real-time signal-processing functions at wide bandwidth (tens of megahertz) making them attractive for use in radar, sonar, and communication equipments. Theoretical results show how specific arrangements of physically realizable SAW chirp filters permit Fourier transformation of both baseband and IF input signals. The operation of SAW Fourier-transform processors, with separate real- and imaginary-baseband inputs, highlights their use as wide-band "FFT-type" processors. The application of individual SAW-based processors to spectrum analysis, network analysis, beamforming, and frequency-hopped wave-form synthesis is discussed and demonstrated. Finally, the use of combinations of two or more SAW Fourier-transform processors for cepstrum analysis and programmable correlation is described.

On the computation of the complex cepstrum

Abstract: A technique based on fitting splines to the phase derivative curve is presented for the efficient and reliable computation of the complex cepstrum. The frequency sampling and the thresholds of the adaptive phase unwrapping algorithm are studied as a function of the radii of the zeros of the signal.

Cepstral synthesis of Japanese from CV syllable parameters

Abstract: Natural and intelligible Japanese was synthesized from consonant-vowel (CV) syllable parameters. Japanese has approximately one hundred CV type syllables which are represented by Japanese alphabet. Each CV syllable is analyzed and stored in the form of cepstrum corresponding to a true log spectral envelope. The speech synthesis system converts character strings into the cepstral parameters with an appropriate dynamic transition from one syllable to the next, and generates a smooth pitch pattern from some pitch informations. In this system, the log magnitude approximation (LMA) filter is employed as parametric synthesizer. The LMA filter generates an acoustic waveform from the cepstrum in real time.

Some notes on phase in speech signals

Abstract: The preservation of phase information in the analysis of speech signals is, quite properly, assuming a role of increased importance. The results of a series of experiments with transformations that destroy speech signal amplitude information but retain phase are reported. These experiments include the preservation of only the phase of a short-term Fourier representation of a signal, preservation of only the phase information of the short-term direct phase cepstrum, and preservation of only the phase of the analytic signal. All of these "phase preserving" transformations are shown to retain many important features of speech signals.

Channel vocoder based on c.c.d. discrete-Fourier-transform processors

Abstract: This paper investigates how charge-coupled-device (c.c.d) modules could be applied in an alternative implementation of the now established channel vocoder. The objective is to demonstrate the potential of c.c.d. technology in the realisation of low-cost low-power speech processors. In this proposed implementation, the central processor is based on a c.c.d. discrete Fourier transform (d.f.t.) algorithm which is multiplexed to provide spectral data for channel compression and cepstral data for pitch detection. Synthetic speech is reconstructed from zero phase impulse responses by a novel convolution technique designed to eliminate frame discontinuities. Extensive computer simulations have been used to determine a suitable hardware architecture and this is compared to a c.c.d. switched-capacitor parallel-filter-bank channel vocoder in terms of performance and total chip count. It is concluded that a very small c.c.d. channel vocoder configuration could be based on a combination of the d.f.t. analyser and the filter-bank synthesiser.

Optimal estimation of time series functions

Abstract: The principal methods of spectral estimation include kernel smoothing, ME-AR methods, and the Prony algorithm. In this paper we consider not only the estimation of spectral density matrices, but also related time series functions such as transfer functions, gain, phase, cepstrum, and so on. Computational algorithms are developed for use of splines as estimators. Smoothing splines are shown to be special cases of kernel smoothers so that properties of kernel smoothers carry over to spline estimators. Optimality of these estimators is discussed.

Queue by Signal Processing Methods

Abstract: AbstmcGTwo methods, based on signal processing techniques, are presented for obtaining numerical solutions for the general single server queue with fvstcome, first-served discipline. The first method is basecl on the use of the fast Fourier transform (FFI’) for producing iterative solutions to a discrete version of Lmdley’s integral equation for both nonsteady state and equilibrium conditions. The second method makes use of the complex cepstrum, implemented with the FFl‘, for providing direct solutions with the queue in equilibrium.

Maximum entropy method spectral estimation applied to power cable diagnostics via cepstrum processing

Abstract: The Cepstrum processing method has been used on power cables (1, 2) for determining regions of damage. The method consists of: (1) observing the spectrum of an original broad band signal source, (2) introducing the signal into the cable, (3) computing the change in the observed spectrum (as a result of echos from regions of cable non-uniformity) when the signal is injected into the cable, and finally (4) computing the power-spectrum of the change in observed spectrum. A limitation of this technique is the use of band-limiting spectrum analyzers. Their limited bandwidth reduces range resolution estimation when used with the Fast Fourier Transform (FFT) technique. The maximum entropy method (MEM) is a more useful spectral estimator for this measurement technique. Examples are presented which show a comparison of the FFT and MEM techniques applied to practical cables.