COMPARISON OF FORMANT TRANSITION BASED STOP CLASSIFIERS: TIME-VARYING AND TIME-INVARIANT SIGNAL MODELS

Abstract

A feature set that captures the dynamics of form ant transitions is utilized to classify the unvoiced stop consonants. The second formant and its slope are used to characterize the transition between the vowel and closure in a VOV environment. The performance of a feature set obtained by means of a time-varying, data-selective model for the signal is compared with that of a standard time-invariant (LPC) and data-selective, time-invariant models. The different feature sets are evaluated on a database consisting of 10 talkers. A two-fold reduction in the error rate is obtained by means of the time-varying model. The performance of three different classifiers is presented. A novel adaptive algorithm, termed Learning Vector Classifier, is compared with standard K-means and LVQ2 classifiers. Speaker independent error rates of 11% are obtained for the 3-way classification task. Further improvements, expected when an expanded time-varying feature set, is coupled with information from the burst, could lead to very high quality stop classification.