ANN Based Disfluent Speech Classification

Abstract

About 1% of theworld population suffer from stuttering, a continued involuntary repetition ofsound; especially initial constants. Another name for this is speech dis-fluency or Disturbed Speech. This includes word repetition, syllable repetition, prolongation, and interjection. The existing algorithm focuses wholly on either extraction or classification. This paper uses "Artificial Neural Network" or ANN and implements automatic analysis of disfluent speech by extracting "Mel frequency cepstral coefficient" or MFCC, Delta MFCC, Delta Delta MFCC and prosodic features like pitch, energy, duration and the like. This paper aims at improving the fluency of the stuttered speech.