A SILENCE REMOVAL AND ENDPOINT DETECTION APPROACH FOR SPEECH PROCESSING

In this paper a brief overview of silence removal and voice activity detection is discussed and a new method for silence removal is suggested. The objective of suggested method is to delete the silence and unvoiced segments from the speech signal which are very useful to increase the performance and accuracy of the system. Endpoint detection is used to remove the DC offset value from the signal after silence removal process. Silence removal and Endpoint detection are main part of many applications such as speaker and speech recognition. The proposed method uses Root Mean Square (RMS) to delete the unvoiced segments from the speech signal. This work showed better results for silence removal and endpoint detection than existing methods. The performance of this research work is evaluated using MATLAB tool and accuracy of 97.2% is achieved.