Comprehensive design and development of time efficiency speaker recognition model from front end to back end

Abstract

The rapid development of the forensic science technologies has been evolved speaker recognition to becoming one of the research topics. However, pattern classification from speech signal remains as challenging problem encountered in general speaker recognition system, including speaker verification and speaker identification. Conventional speaker recognition researches are almost directed towards accuracy problems, not time processing problems. Due to the needs of reduction time processing of speaker recognition system, this research focuses on develop a comprehensive design of speaker recognition model from front end to back end which able to process speaker data in short time limit. In the front end process, we introduce some pre-processing techniques to enhance the speech signal. Whereas, for the back end process, we propose a decision function by using vector quantization techniques to decrease the training model for GMM in order to reduce the processing time. Experimental result shows that our hybrid VQ/GMM method always yielded better improvements in accuracy and bring almost 30% reduce in time processing. In this paper, a new, robust and simplicity computation method of pattern classification technique for speaker identification system is proposed. Consequently, this research is intended to develop a fully optimize ways speaker identification approach from hybrid modeling.