Pitch Synchronous Analysis of Voiced Sounds

Abstract

A study of vowel sounds by means of a spectral analysis keyed synchronously to the voice pitch has been carried out. Spectra are obtained by Fourier analysis of individual pitch periods which were established by visual inspection of oscillograms. A digital computer served as the analyzer. The spectra are represented by a pattern of zeros and poles obtained by a process of successive approximation, again carried out by computer. The contributions from vocal tract and glottal source can be uniquely separated and examined. These results show that vowel sounds can be represented by a sequence of poles arising from the vocal tract and a sequence of zeros charactering the izglottal excitation. The frequencies of the vocal tract poles agreed with previous measurements, but the damping factors were not entirely consistent with earlier estimates. The zeros showed approximately uniform frequency spacing, particularly at high frequencies. A theoretical development indicated that this characteristic was to be expected from the known structure of the glottal excitation. The zero pattern was used to estimate the ratio of open-to-closed time for the glottis during voicing. PEECH is the acoustic result of certain events in the vocal apparatus. Indeed, many of the characteristics of these events can be deduced from the acoustical properties of the speech itself. For instance,it has been long realized that speech formants are a reflection of vocal tract resonances, and that the quasi-periodic nature of vocalized speech can be attributed directly to the vocal cord excitation. Usually, such features have been studied with the aid of the sound spectrograph and other special instruments. Though of great utility, these instruments have limitations, both in the range and flexibility of their analysis parameters, and in the dynamic range and resolution of their output displays.With the advent of digital computers as analysis tools, much more sophisticated processing and display have become possible. 2'a This paper presents a basic speech spectrum analysis 4'5 which takes advantage of the greater resources of the computer. The results indicate that this analysis leads to a much more precise acoustic representation of the speech wave than heretofore feasible. Consequently, this representation can be tied quite accurately to vocal tract events.In the first step of the analysis the speech time waveform is segmented into pitch periods which are then subjected to a Fourier expansion. The resulting spectrum is approximated by a number of resonances (poles) and antiresonances (zeros). Utilizing the assumption that each period is one of an infinite sequence of identical periods, this "pitch synchronous" representation can be related easily and precisely to vocal characteristics. In particular, the physical constraints imposed by the known structure of the vocal tract permit the • This paper is the written version of a talk given at the 58th Publishing Company, Amsterdam, 1960). 179 poles and zeros to be assigned uniquely to either the vocal tract...

WORLD: A Vocoder-Based High-Quality Speech Synthesis System for Real-Time Applications

“…WORLD uses an accurate spectral envelope estimation algorithm called CheapTrick [16], [18]. CheapTrick is based on the idea of pitch synchronous analysis [26] and uses a Hanning window with the length of 3T 0 . First, the power spectrum is calculated on the basis of the windowed waveform.…”

Error Evaluation of an F0-Adaptive Spectral Envelope Estimator in Robustness against the Additive Noise and F0 Error

“…Step: F0-Adaptive Windowing An F0-adaptive Hanning window is used on the basis of pitch synchronous analysis [18]. In CheapTrick, the length of the window is set to three times the fundamental period (T 0 ).…”

PLATINUM: A method to extract excitation signals for voice synthesis system

“…The STRAIGHT [7] and TANDEM-STRAIGHT [8] have been proposed to solve this problem. They use the pitch synchronous analysis [9] to improve the estimation performance of the spectral envelope. Furthermore, aperiodicity is used as the parameter to represent not only the periodic signal but also the aperiodic signal.…”