A Yet Efficient Communication System with Hearing-Impaired People Based on Isolated Words of Arabic Language

摘要

Although the mental capabilities of hearing-impaired people and hearing people are the same, reading and writing skills of those with hearing loss are much lower. Therefore, developing communications systems using sign languages becomes feasible and of interest. On the other hand, despite the fact that Arabic language is currently one of the most common languages worldwide, there has been only a little research on Arabic speech recognition relative to other languages such as English and Japanese. In this paper, a speech recognition system was developed which basically identifies the Arabic words that a hearing person speaks into a microphone to be then translated into a video, implements real Arabic sign language, which is easily understood by a hearing-impaired person. The speech recognition process carried out in this paper involves mainly using voice activity detection (VAD), Mel-frequency cepstral coefficients (MFCC), and dynamic time warping (DTW) algorithms. Moreover, delta and acceleration (delta-delta) coefficients have been added for the reason of improving the recognition accuracy. Utilizing the best set up made for all affected parameters to the aforementioned techniques, the proposed system achieved a recognition rate of about 98.5% which outperformed other relevant hidden Markov model (HMM) and artificial neural network (ANN)-based approaches available in the literature.