Title :
Unsupervised and nonparametric Bayesian classifier for HOS speaker independent HMM based isolated word speech recognition systems
Author :
Zribi, M. ; Saoudi, S. ; Ghorbel, F.
Author_Institution :
Groupe de Recherche Images et Formes, Inst. Nat. des Telecommun., Villeneuve d´´Ascq, France
Abstract :
We consider a speaker independent hidden Markov model (HMM) based isolated word speech recognition system. The most general representation of the probability density function (PDF), in the classical HMM, is a parametric one (i.e., a Gaussian). We derive an unsupervised, nonparametric and multidimensional Bayesian classifier based on the well known orthogonal probability density function (PDF) estimator which does not assume any knowledge of the distribution of the conditional PDFs of each class. Such a result is possible since this nonparametric estimator is suitable and adapted to the expectation maximization (EM) mixture identification algorithm
Keywords :
Bayes methods; hidden Markov models; higher order statistics; nonparametric statistics; probability; spectral analysis; speech recognition; EM mixture identification algorithm; HOS speaker independent HMM; LSP coefficients; PDF estimator; conditional PDF; expectation maximization mixture identification algorithm; hidden Markov model; isolated word speech recognition systems; line spectrum pair; multidimensional Bayesian classifier; nonparametric Bayesian classifier; nonparametric estimator; orthogonal probability density function; unsupervised Bayesian classifier; Bayesian methods; Cepstral analysis; Cepstrum; Hidden Markov models; Multidimensional systems; Performance analysis; Probability density function; Speech recognition; Telecommunications; Vocabulary;
Conference_Titel :
Statistical Signal and Array Processing, 1996. Proceedings., 8th IEEE Signal Processing Workshop on (Cat. No.96TB10004
Conference_Location :
Corfu
Print_ISBN :
0-8186-7576-4
DOI :
10.1109/SSAP.1996.534850