Showing papers on "Cepstrum published in 1978"
Dissertation•
01 Jan 1978
TL;DR: This paper presents a meta-modelling system that automates the very labor-intensive and therefore time-heavy and therefore expensive and expensive process of computer programming called “hacking”.
Abstract: Thesis. 1978. Elec.E.--Massachusetts Institute of Technology. Dept. of Electrical Engineering and Computer Science.
26 citations
10 Apr 1978
TL;DR: By using the log magnitude approximation filter as the speech synthesizer, a low bit rate cepstral vocoder was obtained and this vocoder can provide a better spectral fit than that provided by all-pole model.
Abstract: By using the log magnitude approximation filter as the speech synthesizer, a low bit rate cepstral vocoder was obtained. The synthesis filter is of minimum phase and of pole-zero type, and this vocoder can provide a better spectral fit than that provided by all-pole model. The filter can be easily obtained without transforming the cepstrum to an impulse response and the computation for the filter coefficients is non-iterative. The spectral envelope information is transmitted in a form of cepstral values. Since the coefficient sensitivities of the synthesis filter is sufficiently small, a fairly coarse quantization of the cepstrum is allowed. A reasonable speech quality vocoder was realized at the overall bit rate of 1770 bits/s.
13 citations
TL;DR: In this paper, an analysis of homomorphic deconvolution for processing the output of a distortional N-path channel is presented, where the interference generated by the various convolutions of the channel response with itself is traced to the interference produced by the interference.
13 citations
TL;DR: In this article, an elemental filter with a component of the quefrency response is presented, and the desired log magnitude and phase responses are realized by a cascade connection of the elemental filters.
Abstract: The desired log magnitude and phase responses are realized by approximating the log frequency response (the complex logarithm of the frequency response). Since the log frequency response is represented by its Fourier series in terms of the quefrency response (the cepstrum of the impulse response), the desired log frequency response is approximated by the filter with a finite length quefrency response. The elemental filter presented in this paper has a component of the quefrency response. Therefore, the quefrency response is realized by a cascade connection of the elemental filters for all its quefrency components. The cascade filter provides the best mean-square approximation to the desired log frequency response. By introducing the elemental filter, digital filters can be designed in the quefrency domain. By designing digital filters in the quefrency domain, the best approximation filter for the desired log frequency response can be easily realized.
6 citations
6 citations
01 Apr 1978
TL;DR: This paper describes the use of comb lifter in the Cepstrum analysis which is usefull for extraction of formant frequencies and a new type of adaptive lifter is presented which does not need the pitch detection and is useful in case of connected speech.
Abstract: This paper describes the use of comb lifter in the Cepstrum analysis which is usefull for extraction of formant frequencies. Good result is obtained with Han lifter in case of low pitch frequency. But, in case of high pitch frequency, the separation of two formants is sometimes impossible. The comb lifter is presented to overcome this difficulty, that is, the length of the lifter is fixed so as to separate the closest formants, and the peaks in the cepstrum due to the periodicity of speech are suppressed by using comb lifter. Successful results have been obtained in the experiments using two types of comb lifter. A new type of adaptive lifter is also presented which does not need the pitch detection and is useful in case of connected speech.
4 citations
TL;DR: In this article, a statistical analysis of the expected value of the center position of a discrete random noise process is presented, and it is shown that for processes that can be described by the convolution of a white series and a disturbance wavelet, the centre position is independent of the phase property of the wavelet.
Abstract: A finite realization of a discrete random noise process may be considered as a one-sided energy signal. Its phase property can then be described by means of the center position. The samples of such a realization are the components of a random signal vector and the center position is therefore a random variable. A statistical analysis shows that the expected value of the center position equals half the time duration of the realization. This implies that the Z-transform of the realization may be expected to have an equal number of poles and zeros inside and outside the unit circle. The standard deviation from the expected value of the center position is shown to depend on the time duration of the realization and on the autocorrelation of the process. It follows that, for processes that can be described by the convolution of a white series and a disturbance wavelet, the center position is independent of the phase property of the wavelet. A conclusion based on these results is that the homomorphic technique of wavelet estimation through cepstrum stacking must give questionable outcomes. Another conclusion is that the super-position of a realization of random noise on a minimum phase wavelet will in general give a mixed phase resulting signal. It is pointed out that schemes for the derivation of deconvolution filters do not take account of this phenomenon.
2 citations
10 Apr 1978
TL;DR: A new technique for constructing representations of signals belonging to a class that includes arterial blood pressure waves and voiced speech is described, called cepstral inverse filtering, which optimizes the pole and zero locations of a discrete-time system that models the signal production process.
Abstract: This paper describes a new technique for constructing representations of signals belonging to a class that includes arterial blood pressure waves and voiced speech. Called cepstral inverse filtering, it optimizes the pole and zero locations of a discrete-time system that models the signal production process. These locations are chosen so that the cepstrum of the model impulse response is a least-squares estimate of that of the signal under analysis. The iterative procedure employed in solving this non-linear estimation problem is made computationally efficient through the use of simple digital filtering techniques.
2 citations
01 Apr 1978
TL;DR: In this article, a pusle-shift P-M generated combined wave model is used to decompose the combined wave for which the reflected wave frequency information is altered by the structure and no longer is the same as the incident wave, such as occurs with shorelines and many breakwaters.
Abstract: : Cepstral analysis decomposition techniques work extremely well with the pusle-shift P-M generated combined wave model and not so well with the forward-shift P-M generated combined wave model. Problems exist with identification of delay time as well as with decomposition of the combined wave when cepstral analysis techniques are applied to P-M generated wave-models which vary tau and/or the reflection coefficient, alpha, as a function of frequency. More research is required before this decomposition technique can be successfully applied to real-world ocean waves. It may never work for decomposition of combined waves for which the reflected wave frequency information is altered by the structure and no longer is the same as the incident wave, such as occurs with shorelines and many breakwaters. Of the three cepstrum types examined, the power cepstrum is the best indicator of delay time between incident and reflected waves. The phase cepstrum does not seem to offer much information other than reinforcement or possible validation of the power cepstrum information.
1 citations
01 Oct 1978
TL;DR: By using a log spectrum approximation filter modeling a true log spectral envelope, a low bit rate and high‐quality vocoder was realized and the speech synthesis is not based on the homomorphic method.
Abstract: By using a log spectrum approximation filter modeling a true log spectral envelope, a low bit rate and high‐quality vocoder was realized. The true log spectral envelope is represented by the low‐quefrency portion of the cepstrum modified by the low‐quefrency components involved in the rectified fine structure of the log spectrum. It can be obtained through an iteration of DFT‐rectifying‐IDFT for the higher quefrency portion of the speech cepstrum. In this vocoder, the speech synthesis is not based on the homomorphic method. The synthesis filter can be obtained without transforming the cepstrum to an impulse response, and the computation for the filter coefficients is very simple. The spectral envelope information is transmitted in the form of a differential cepstrum corresponding to the difference of two adjacent log spectral envelopes, and the data rate is very low.
01 Apr 1978
TL;DR: An apparently new synthesis method is described, windowed synthesis, which circumvents the problem of choosing initial conditions in synthesis using a filter model, and is also applicable to cepstral vocoding.
Abstract: We describe the implementation of a practical system for pole-zero analysis-synthesis which produces synthesized speech that sounds qualitatively different from linear prediction at points of mouth closure. We also describe an apparently new synthesis method, windowed synthesis, which circumvents the problem of choosing initial conditions in synthesis using a filter model, and is also applicable to cepstral vocoding.
27 Oct 1978
TL;DR: A numerical algorithm for performing the spectral factorization of multidimensional spectral density functions and the resulting factors are stable and realizable (i.e., recursible).
Abstract: In this paper, we present a procedure for the spectral factorization of multidimensional spectral density functions. Properties of the multidimensional cepstrum are developed and used as a basis for the procedure. In analogy with Wiener's one-dimensional factorization, the resulting factors are stable and realizable (i.e., recursible). A numerical algorithm for performing the factorization is described, along with its use in obtaining unilateral representations of multidimensional random fields.
01 Apr 1978
TL;DR: In this article, the Comb Lifter method, filtering in cepstrum domain, and the Moving Average method, with variable averaging points, were used to reduce reflections from the floor, walls and ceiling.
Abstract: In this paper, free-field measurements (frequency responses and directional characteristics) in a normal room are described. Two methods for reducing reflected waves are investigated. The Comb Lifter method, filtering in cepstrum domain, and the Moving Average method, with variable averaging points, are described. Approximate free-field responses in full audio frequency band are obtained with reduced reflections from the floor, walls and ceiling.
TL;DR: The signal theory of speech and the possibilities of bandwidth reduction in speech are reviewed and good quality speech has been realised at 9600 bit/s using Adaptive A/D conversion techniques.
Abstract: The paper reviews the signal theory of speech and the possibilities of bandwidth reduction in speech. The current techniques using Vocoders and other methods have been discussed. The channel Vocoders of Dudley and other claim a bandwidth reduction of 15: 1, the Formant tracking Vocoders 10: 1 and Time domain Vocoders give a reduction of 8: 1. Among recent developments, the high fidelity Vocoders compress a 100 Hz-10 KHz speech to a 330–2700 Hz band, suitable for transmission in telephone channels. The Time Domain speech compression equipment using CEPSTRUM technique, realizes a good quality speech at 2400 Hz bit/s, but the system is very complex. Using Adaptive A/D conversion techniques good quality speech has been realised at 9600 bit/s.For computerized signal processing, synthetic speech and speech from printed texts have been developed. Few such techniques have been discussed. The speech recognition problems, however, are still very complex and yet to be formalised.