Continuous Natural Voice Text-independent Speaker Recognition And Dsp-based

Speaker recognition technology belongs to biometrics recognition. Speaker recognition is the process of recognizing who is speaking on the basis of individual information included in speech signals. Compared with other kinds of biometrics technologies, speaker recognition has many advantages: convenient, natural, accurate, economical and well expansibility. Speaker recognition plays an important role in many areas, and has extensive applied foreground.Speaker recognition can be divided into Speaker Identification and Speaker Verification. Depending on the level of user cooperation and control in an application, the speech used for these tasks can be either text-dependent or text-independent. This dissertation develops a DSP-based speaker recognition system, which can be divided into three parts: software processing platform, hardware processing platform and control panel.In the process of speaker recognition, the recognition effect is not well enough when simply use one kind of speech feature. It is necessary to combine many parameters to enhance the recognition effect. But because most fusion techniques require complex arithmetic, in this paper we propose a new fusion technique, termed Information Entropy Fusion. Experimental results indicate that the system has higher accuracy and better validity.This paper introduces the implementation of channel compensation on speaker recognition, compares several different approaches to channel compensation and prospects the future developmental direction of channel compensation.This dissertation develops a high-speed DSP platform using the TMS320C6701 digital signal processor applied to speaker recognition. The DSP platform has strong processing ability, compact structure, and high efficiency. The design idea and structure form has common and practical values.