Automatic Isolated Arabic Speech Recognition and Its Transformation into Signs

In this technological era, providing a decent social integration of the mute communities or for the people with special needs still stands as a challenge. Currently, Sign Language (SL) is the main tool of communication between literate mute communities. About 18 million of the mute community is from the Arabic world [1]. In this framework, this work focuses on the design and development of an effective approach for automatic isolated Arabic speech based message recognition. The objective is to achieve an effective solution with a high level of precision. It is realizable by smartly using the hybrid features extraction and the robust classification techniques. The incoming speech segment is enhanced by the application of appropriate pre-conditioning. The Mel-Frequency Cepstral Coefficients (MFCCs) and the Perceptive Linear Prediction Coding Coefficients (PLPCC) are extracted from the enhanced speech segment. Later specifically designed voting based robust classifier issued to compare these extracted features with the reference templates. The comparison outcomes are the basis of classification decisions. The classification decision is transformed into systematic visual signs. The system functionality is tested with the help of a prototype realization. An average subject dependent Arabic isolated speech recognition accuracy of 92.6% is achieved.