Recognition of Vowel Segments in Spanish Esophageal Speech Using Hidden Markov Models

This paper presents a system for recognition of voiced segments in Spanish esophageal speech. It exposes different algorithms for the feature extraction of speech segment like formant analysis, linear prediction coefficients (LPC) and mel frequency cepstral coefficients (MFCC), as well as, the recognition stage through the hidden Markov models. Simulation results are presented for normal and esophageal speech. The system was implemented in a real time processing platform based on a digital signal processor TMS320C5416 of Texas Instruments