Recognition of intervocalic stops in continuous speech using context-dependent HMMs

Abstract

In this work the design and evaluation of the recognition performance of context dependentHidden Markov Models (HMMs) for the intervocalic voiced and unvoiced stops is described. The phoneme HMMs are context-dependent in order to account for coarticulatory effects. Continuous probability density functions are used for the out putvectors.Initial model parameter estimates are obtained by means of an automatic segmentation procedure for careful modeling of relevant phonetic features. The model structure and the training scheme are directed to associate the most acoustically dis criminativesegments of the consonants with a sequence of states in every consonant model. The speech data base consisted of a total of 2,592 productions of the Spanish Stops /p, t, k, b, d, g/ in intervocalic positions with the vowels /a, i, u/ embedded in VCVCVCV nonsense utterances. The speech data has been produced by two male Argentine Spanish speakers. Phoneme recognition is accomplished finding the state sequence with highest likelihood in an ergodic model formed by the linking of all the context-dependent phoneme models allowing only the phonotactically valid state transi tions.A comparative study of the recognition performance under different degrees of context dependence, and the alternative use of spectral dynamic and energy related pa rametersis presented.

Training HMMs using a minimum recognition error approach

“…The speech data consisted of 432 such utterances per speaker from two speakers. Details of the basic implementation are given elsewhere [8]. The speech signal was sampled at 10 kHz., preemphasized and windowed with a 256-point Hamming window.…”

A method of discrimination and scoring of aspirated and unaspirated sounds for Chinese spoken language learning system

“…In Japanese, the existing methods for discriminating the explosive consonants can be roughly divided into two kinds: discrimination by the spectral structure near the explosive point [4], and discrimination by spectral transitions using articulative coupling [5]. The cepstrum coefficient [6,7] and the critical band spectrum [4,8] are often utilized as feature quantities. Other investigations have dealt with such topics as recognition of the explosive consonants of Japanese by means of the local peaks and slopes of the spectrum [9] and recognition of the voiceless unaspirated explosives and the zero initial of Chinese by using the cepstrum of the explosive part and that of the transition part [10].…”