Study On Speaker Recognition And Implementation On The DSP

Speaker recognition, which is also called voice print, is a technology that attempts to recognize a speaker through measurements of the specifically individual characteristics arising in speaker's voice. Recent years, in the biological identification technologies, speaker recognition technology with unique advantages such as convenience, economy and accuracy is paid much attention to. In the field of information security, applications using speaker recognition technology is gradually increasing,and it will becoming an important and popular security certification method of daily life and work.The thesis first briefly reviews the history of speaker recognition_and then discusses the main theory of it. Feature extraction model, pattern matching model and decision model as the key factors of speaker recognition, several models and algorithm for each part are discussed in detail.The aim of this thesis is to design a practical speaker recognition system based on DSP, in which performance and complexity is relatively the key factors. Based on this conception, according to the compare of time and space consumption for each algorithm, MFCC and VQ are finally chosen to be implemented. Then the recognition performance on various system parameters is emulated to achieve best scheme. Finally the scheme is implemented on the platform of TMS320 C5402DSK system. The main effort then is paid to enhance recognition rate and reduce time and space consume. After testing, system operates correctly, and the target is reached.