Design And Implementation On Text-Dependent Speaker Recognition System

Speaker recognition is a kind of technology which extracts features that can reflect the characteristics of the differences of different speaker voice systems from the speech signal and then finish identification by using pattern classification methods. The speaker recognition is a biometric authentication technology, due to it’s several advantages, it has been more and more frequently applied in the fields of speaker monitor, identity verification, financial security and justice of criminal investigation and other fields, it is also the hotspot in the current voice signal identify areas. In text-dependent speaker recognition system, usually defines the contents of the speech, thereby limiting its application scenarios.This paper focuses on text-dependent and open-set speaker recognition, creates the text-dependent and open-set speaker recognition system, and analyzes the performance of the system. At the same time, this paper also has carried on the preliminary study on the keyword spotting recognition, compared the effects on between modeling by keywords and modeling by phonemes and between long keywords and short keywords in terms of endpoint detection.For the text-dependent and open-set speaker recognition system, this paper first introduces the key technologies of speaker recognition system. Then the paper introduced the practice about some of the key technologies, tested the accuracy rate of identification and verification in this system, tested the anti-noise performance when adding noise to testing speech.For keyword recognition system, this paper first introduces general situation of the keyword recognition, and then introduces the key technology in the field of identifying keywords. Secondly, this paper introduces the construction and implementation of keywords spotting system. Finally, through the experiments, this paper compares long keywords with short keywords about the performance on endpoint detection and keyword verification.