Online Dialogue Voice-based Speaker Recognition Technology Research

Speaker recognition technology is a more effective biometric authentication technology. It analyzes the voice to determine who and when speaks. And It has very broad applications. Existing applications are mostly simple speaker verification and recognition system. In this paper, a speaker recognition technology based conversational speech is proposed, and implementation of the speaker recognition system. Main tasks are summarized as follows:(1) A two-stage BIC segmentation method is proposed, the first phase of the larger window length partitioning strategies to ensure on-line real-time identification method, the second phase of BIC-based segmentation criteria to ensure the accuracy of the speaker segmentation.(2) A VQ-based speaker recognition system is implemented, which aims to improve the traditional LBG algorithm, and to meet the system on-line identification feature better.(3) A differences subspace method in speaker recognition is suggested. It aims to establish the corresponding subspace for each speaker, and to select the highest score of comparing the test-speech to the train-speech as the result. This method improves the system recognition rate accurately. And it reduces the systems'time costing, the system is suitable for on-line identification system uses.(4) An on-line speaker recognition system based conversational speech is implemented at last.