Vector Quantization for Stuttered Speech Recognition

Abstract

In current trends main challenges in stuttered speech recognition are detection of stuttering fluency disorders like repetitions, prolongations, and interjections of words, silent pauses, broken words, incomplete phrases and revisions between the words. In the proposed work, silent pauses and stopgaps of stuttered speech is considered for recognition. Mel Frequency Cepstral Coefficient (MFCC) and K-Means are the anticipated methods for classification of stuttered speech with salient pause and repetitions.