Hybrid SVM/HMM Model for the Arab Phonemes Recognition
Elyes Zarrouk and Yassine Benayed
Multimedia Information System and Advanced Computing Laboratory, Sfax University, Tunisia
Abstract: Hidden Markov Models (HMM) are currently widely used in Automatic Speech Recognition (ASR) as being the most effective models. Yet, they sometimes pose some problems of discrimination. The hybridization of Artificial Neural Networks(ANN) in particular Multi Layer Perceptrons (MLP) with HMM is a promising technique to overcome these limitations. In order to ameliorate results of recognition system, we use Support Vector Machines (SVM) witch characterized by a high predictive power and discrimination. The incorporation of SVM with HMM brings into existence of the new system of ASR. So,
by using 2800 occurrences of Arabic phonemes, this work arises a comparative study of our acknowledgment system of it as the following: The use of especially the HMM standards lead to a recognition rate of 66.98%. Also, with the hybrid system MLP/HMM we succeed in achieving the value of 73.78%. Moreover, our proposed system SVM/HMM realizes the best performances, whereby, we achieve 75.8% as a recognition frequency.
Keywords: ASR, Hybrid System, HMM, MLP, SVM.
Received July 24, 2012; accepted September 27, 2012