Automatic Arabic Speech Recognition- A Comparative Study

Abstract

The cutting-edge studies on Automatic Speech Recognition approach have reported exceptional accuracy rates that are even comparable to human transcribersposing a question if machine has reached human performance. Automatic Speech Recognition can be used as a biometric authentication technique, which is essential in ciphering many applications used. In light of the Arabic language, only few studies have proposed to assess the effectiveness of using Automatic Speech Recognition in Arabic language; therefore, this study aims to implement Arabic speaker recognition using three different algorithms, including (i) Dynamic Time Warping (DTW), (ii) Gaussian mixture model (GMM), and (iii) Support Vector Machine (SVM). To measure the effectiveness of these algorithm in recognizing the Arabic speech, two datasets are used to train and test them, which are: (i) speech agent archive, and (ii) Arabic speech corpus. The results reveled that the DTW outperforms the GMM and SVM in terms of accuracy, precision, recall and fmeasure, as it achieves 95.7%, 96%, and 95%, and 96%, respectively.