A Novel Phonemes Classification Method Using Fuzzy Logic
Ines Ben Fredj · Science Journal of Circuits Systems and Signal Processing · 2013
In this study, we will interest in phonemes classification of Timit database using Fuzzy Logic. The fuzzy method consists in the extraction of a three fuzzy-reference vectors: maximal, mean and minimal. To classify a phoneme request, we calculate its degree of membership to all defined classes. The class of a phoneme request is, then, the one which maximizes one degree of membership calculated according to reference vectors. Different techniques of speech analysis are used for evaluation. Results show that fuzzy logic can provide a significant issue when mathematical rigor is not present as to the signal processing since the retained recognition rates was 90,85%, 22,96%, 98,57% and 91,73% for respectively MFCC, LPC, PLP and RASTA PLP.