Study of speech analysis techniques for the phonemes classification using fuzzy logic

In this work, we study speech analysis techniques to classify phoneme using a method of fuzzy logic. The used techniques are Mel Frequency Cepstral Coefficient (MFCC), Perceptual Linear Prediction (PLP) and RelAtive SpecTrAl-Perceptual Linear Prediction (RASTA-PLP). The fuzzy logic method is characterized by three fuzzy reference vectors: the maximal vector, the mean vector and the minimal vector. To classify a phoneme request, we calculate the degree of membership of this phoneme to all classed of the base of phonemes. The class of phoneme request is them the one which maximizes one degree of membership calculated according to reference vectors. For evaluation, a comparative study was operated to fix on the most perfect features extraction technique used.