Unsupervised and Non Parametric Bayesian Classifier for HOS Speaker Independent HMM Based Isolated Word Speech Recognition Systems

  • Authors:

  • M. Zribi;S. Saoudi;F. Ghorbel

  • Affiliations:

  • -;-;-

  • Venue:
  • SSAP '96 Proceedings of the 8th IEEE Signal Processing Workshop on Statistical Signal and Array Processing (SSAP '96)
  • Year:
  • 1996

Quantified Score

Hi-index 0.00

Visualization

Abstract

Here, we consider a speaker independent Hidden Markov Model (HMM) based isolated word speech recognition system. The most general representation of the probability density function (pdt), in the classical HMM, is a parametric one (i.e, a Gaussian). We intend here to derive an unsupervised, non parametric 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 result becomes possible since this non parametric estimator is suitable and adapted to Expectation Maximization (EM) mixture identification algorithm.