Automatic Arabic recognition system based on support vector machines (SVMs)

Automatic Speech Recognition (ASR) for Arabic word has been developed in this work. The system has the ability to recognize word that is uttered by the speaker. In this paper, an approach using support vector machines (SVMs) for identifying Arabic word based on the speaker speech is proposed. The proposed SVMs-based Automatic Speech Recognition system is tested experimentally using words uttered by 20 native Arabic speakers. The Mel Frequency Cepstral Coefficient (MFCC) is adopted as a feature and later used as an input to the SVM-based identifier. The performance of the proposed technique has been investigated, especially for multiclass classification and it is found to produce good accuracy within short duration training time.