Abstract:

Consonant recognition has important clinical significance in the assessment of dysarthria, while the consonants are  so short and unstable that the recognition results of the traditional methods are ineffective. The algorithm described in this paper extracts a new feature(DWTMFCCT) of the consonants employing wavelet transformation. And the difference of similar consonants can be described more accurately by the feature. And then the algorithm classifies consonants using a multiclass fuzzy support vector machine(FSVM). In order to reduce the computation complexity caused by using the standard fuzzy support vector machines for multiclass classification, this paper proposes an algorithm based on two stages. The experimental results show that the proposed algorithm can get better classification results while reducing the training time greatly.

Key words: consonant recognition;fuzzy support vector machine;wavelet transform;Mel frequency