Voice Waveform Research Articles

An Algorithm for Voice Waveform VQ Codebook Design Based on PCA

- Speech compression plays an important role in applications which require the minimization of the storage and/or transmission requirements, such as multimedia, integrated services digital networks (ISDN), voice response systems and cellular telephony. Vector quantization (VQ) is a well-known compression technique which has been widely used in many speech coding systems. In the present work, an algorithm for designing codebooks for voice waveform vector quantization is presented. Simulations are carried out to compare the performance of the proposed algorithm, which is based on principal component analysis (PCA), to the performance of two other design approaches: the traditional LBG (Linde-Buzo-Gray) algorithm and a self-organizing neural network algorithm.

Method for processing speech signal in speech processing system

A method for processing an input speech signal to be applied to a CELP vocoder has the steps of obtaining preliminary pitch search intervals by a preprocessing autocorrelation expression from a pitch lag of a synthesized speech signal which is synthesized from a residual signal of the input speech signal; computing coefficients of pitch filter with respect to the preliminary pitches; searching a high interval in the autocorrelation; and removing the remaining interval other than the high interval in the pitch lag. Since the present invention proposes a speech processing method which uses only a high interval in autocorrelation of a voice waveform in pitch-searching, and where such a speech processing method is embodied in a CELP vocoder, total computation time of the CELP vocoder can be decreased 37% or more without lowering speech quality. Therefore, a digital signal processor, which is low in price and is slow in speed, can be embodied in a CELP vocoder.

Speech analysis apparatus for extracting glottal source parameters and formant parameters

An LPC analysis of a voice waveform is performed by an LPC analyzing circuit before analyzing vocal tract parameters by a vocal tract parameter analyzing circuit. The result of the analysis is converted into the components of source parameters by converting circuits and to eliminate the component of the source parameters from the voice waveform by an inverse filter. The components of the vocal tract parameters obtained by the analysis in the vocal tract parameter analyzing circuit is eliminated from the voice waveform through an inverse filter.

声の変動特性

The variability and controllability of vocal fundamental frequency (F0) are investigated by analyzing the acoustic and perceptual characteristics of sustained vowel samples recorded from patients with spasmodic dysphonia (SPD) as well as patients with various types of laryngeal disorders and normal controls. All the pathological groups show larger variations in F0 or lower controllability than the normal controls. The voice segments perceptually judged as non-sporadic show significantly larger variations in F0 than the normal controls. From the voice segments perceptually judged as sporadic, instantaneous changes in F0 or in amplitude, chaotic changes in voice waveform such as bifurcations are observed. Based on the obtained results, a model of normal and disordered voice control mechanism is proposed, in which morphological and neural abnormalities affect voice variability. The acoustic analysis method described here is found beneficial in analyzing voice variability and controllability, and can provide useful information to know normal and disordered voice control mechanisms.

Voice coding apparatus

VOICE CODING APPARATUS ABSTRACT OF THE DISCLOSURE A voice coding apparatus includes a pitch detecting circuit which detects a pitch period of a voice signal; a pitch waveform generating circuit which samples the voice signal for a plurality of pitches based on the pitch period detected by the pitch detecting circuit and which generates a waveform of one pitch from the waveform of the plurality of pitches; a band restriction circuit which restricts the frequency band of the one pitch waveform generated in the pitch waveform generating circuit; and a coding circuit for coding the voice waveform which is band restricted in the band restriction circuit. The sampling number of the waveform for a plurality of pitches and the restricted bandwidth can be changed in accordance with the amount of the pitch period extracted in the pitch detecting circuit. Further, the pitch detecting circuit is able to correctly detect the pitch period even when the pitch period is not a multiple of the sampling period.

Primary Tracheojejunal Shunt Operation for Voice Restoration following Pharyngolaryngoesophagectomy

A primary tracheojejunal shunt operation was performed for voice restoration following pharyngolaryngoesophagectomy with free jejunum reconstruction for advanced hypopharyngeal cancer. A fistula was created between the membranous part of the trachea and the lower part of the transplanted jejunum. The membranous part of the trachea was tubed to construct the tracheojejunal shunt. All three patients who had the tracheojejunal shunt operation retained phonatory function. Pitch formation was seen in the voice waveform with use of the tracheojejunal shunt. No leakage was seen at all during deglutition and a swallowing function was obtained in all patients who had the tracheojejunal shunt.

さ声の質と量の客観的評価法

A spectrogram of a normal sustained vowel, taken with a narrow filter, shows welldeveloped harmonics as equally spread horizontal stripes, while that of a hoarse voice presents the noise components as a cloudy shadow between harmonics. A small laboratory computer was used to evaluate hoarseness objectively. Three pitch periods extracted from the voice waveform were analized at up to 5.5KHz harmonic frequency by Fourier Series Expansion. Since it offers a dispersal spectrum with good frequency resolution, the acoustic energy of the harmonics is calculated separately from that of the noise components. This study was carried out in 28 normal subjects and 30 patients. The ratio of harmonics to noise (Ra) of the normal group ranged from 15.0 to 23.5dB, and in 95% of normal cases the Ra value renged from 15.4 to 23.3dB. With this criterion, pathologic voices could easily be differentiated from normal. The Ra value also showed a definite correlation with the auditory impression and the visual impression of the spectrogram, suggesting that this is a useful tool in quantitative evaluation of the degree of hoarseness.Furthermore, the relation between the distribution of the noise energy and the voice quality in each case was investigated. The ratio of harmonics to low frequency noise components below 1.38KHz (RL) corresponded well with the R factor (roughness) of the auditory impression, while that of harmonics to high frequency noise components above 1.38KHz (RH) correlated well with the B factor (breathiness). More detailed investigations of noise energy would probably increase the success of differentiation among various laryngeal diseases.

A Novel Digital Speech Detector for Improving Effective Satellite Capacity

A probabilistic analysis has been done by comparing cumulative distributions of voice and typical telephone-line noise, and the characteristics of voice waveform have been investigated in order to determine a new criterion for distinguishing the two signals. This digital speech detector is extremely sensitive to the onset of voice signal but much less sensitive to typical telephoneline noise.

Digital Deresonation of Acoustic Recordings

Recent research into the problem of eliminating blur and reverberation from digital signals has led to the invention of a statistical deconvolution method based on the theory of homomorphic filtering. The application described here involved the removal of recording horn resonances from old acoustic disks. A major distortion in these disks is the convolution of the recording horn impulse response with the voice waveform. This impulse response is estimated by chopping the recorded signal into many 1-sec intervals and averaging logarithms of the corresponding Fourier transforms. After averaging several such segments, the impulse response emerges in the form of its log Fourier transform and is thereafter removed in a digital convolution involving a computed system inverse. The resulting enhanced sound is generally considered far more acceptable and informative than the original although a slight increase in apparent surface noise is induced. A demonstration will be made. [This research was supported in part by the University of Utah Computer Science Division and by the Advanced Research Projects Agency of the Department of Defense, monitored by Rome Air Development Center, Griffiss, AFB, New York 13440, under contract.]