期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2019
卷号：19
期号：3
页码：128-133
出版社：International Journal of Computer Science and Network Security
摘要：In the automatic speech recognition system, several techniques of feature extraction have been studied at different values of signal-to-noise ratio. This paper suggests to develop a new approach of the speech signal such as MFCC-RPCA in order to obtain a higher recognition rate. Thus, MFCC is one of the most commonly used features for speech recognition systems. In previous years, the research on robust principal component analysis (RPCA) has been attracting much attention, in many domains, such as image processing, separation of music/voice, etc. The purpose of this paper is based on the separation speech/noise. In the experimental part, isolated words were chosen from TIMIT database, under additive impulsive and convolutive noise conditions, with SNR (signal to-noise-ratio) ranges from -3 to more than +9 db using the HTK platform (Hidden Markov Model Toolkit). Experimental results have shown that the proposed method has enhanced the quality of signals by reducing the noise level.
