Arabic Continuous Speech Recognition Based on Hybrid SVM/HMM Model
Elyes Zarrouk, Yassine Ben Ayed, Faı̈ez Gargouri · Communication and Signal Processing · 2018
The Hidden Markov Models (HMM) achieved a huge progress, but imperfectly they still suffer from their lack of discrimination capability especially on speech recognition. Therefore, as a way to improve results of recognition systems, we engage Support Vectors Machine (SVM) which works like an estimator of posterior probabilities, in as much as they are characterized by an immense discriminatiom and a big predictive power. Moreover, they are based on structural risk minimization (SRM) where the goal is to learn and gain a classifier that can minimizes a bound on the expected risk, rather than the empirical risk. In this work we present a new approach for automatic labeling with respect to the syntax and the grammar rules of Arabic language. The obtained results for the Arabic speech recognition system based on triphones are 64.68% with HMMs standards and we achieve 76.96% as best recognition rate of a tested speaker with the proposed system SVM/HMM. Consequently, the WER obtained for the recognition of continuous speech by the three systems proves the performance and the powerfulness of SVM/HMM. The speech recognizer was evaluated with ARABIC_DB corpus and performs at 11.42% WER as compared to 13.32% with triphones mixture-Gaussian HMM system.