Design And Establishment Of Anti-Recording Playback Speaker Recognition System

Biometrics are more and more used in identity because of their good safety. Biometric technologies are to use the human oneself s physiology or behavior characteristics that can not be duplicated as features of identity, Human biological features include fingerprints, voice, face, retina, iris, palm, palm vein, skeleton, shape, etc. The speaker recognition is of no need of touch, of simple requirement for equipment, of high efficiency for processing and of high reliability, it coulde be widely used in Entrance guard system, attendance system, judicial investigation, remote identity authentication, etc. At the same time, in the practical application, the speaker recognition is faced with the problem of counterfeit speech attack, it is very easy to record the speaker’s voice and playback to attack the speaker recognition. Therefore, the development of Anti-Playback function in the speaker recognition system is very necessary and practically meaningful. This paper’s work are as follows.1. Discusses the related theory and technology of speaker recognition and anti-playback, Presents the design scheme of speaker recognition system with anti-playback function.2. Design and build Speech Database including original and playback speeches, in order to gain full and reliable reliable experimental data to verify the anti-playback research.3. Design and Establish the Speaker Recognition System with Anti-Playback function through MFC/C++coding. The system is devide into3modules, they are Features Extraction Module, Training Models Module and Recognition Test Module; These three modules are functional indepent, although they constitute the general feature database and model database; Each module supplies flexible function interfaces for the reuse and further development of the codes; Using MFC technology, the system has the visualization, simple easy operation characteristic because of the simple and clear man-machine interface.4.Overall assessment of the performance of the system, it includes operation efficiency evaluation and identification effect evaluation, provides reliable experimental data for the research of anti-playback of speaker recognition, and also good simulation and reference for the construction of the online system.