Design Of Speaker Recognition System Based On FPGA

Identity authentication is used everywhere in our life in the information age, however, traditional authentication ways is lost easily, damaged easily, broken easily, unportable, etc. the speaker identity authentication technology based on voiceprint recognition, voice messages contain voiceprint access convenient and naturally, the authentication process is less invasive, certification is low cost, suitable for remote authentication, which makes the speaker recognition technology has important research value and great social application value. Voiceprint recognition technology is now mainly implemented by software, the speed is slow and reliability is poor, SOPC(System on a Programmable Chip) is a software and hardware combined system based on FPGA(Field Programmable Gate Array), the advantages of good reliability, fast operation, easily upgrade and low cost, make FPGA-based SOPC technology as a better way to achieve speaker recognition system.The design is aimed at the research on a area of Heilongjiang provence human resources and social security bureau’s pension retirees authentication project every year, the area has hundreds thousands of the retired personnel, they have to confirm retirement personnel’s identity every year, prevent the parties have passed away, their relatives do not report continuing receive the pension, but a lot of the old peoples because of age, illness or recuperate in the far coastal regions and other reasons, come to the human resources and social security bureau confirm identity face to face is really inconvenient, thus the human resources and social security bureau proposed the identity authentication project based on speaker recognition.Through research on the voice pattern characteristic parameters and pattern matching method research status at home and abroad, analysis the corresponding recognition rate and FPGA realizability, choice the MFCC(Moore frequency cepstral coefficients) for feature parameters, DTW(dynamic time warping) method for pattern matching mode, The hardware platform choice DE2 of Altera company. The main work of this paper include improved the algorithm of MFCC feature parameter extraction and DTW recognition pattern, verified the algorithm by MATLAB; with pro-processing, MFCC extraction, voiceprint library establish and DTW pattern recognition for speaker aspeech signal in EP2C35 chip, achieved the speaker recognition system; finally, integrated the system, download the bit file, debug and analysis the experimental results.Test the function of system in a sample of 10 individuals, the average recognition rate of system is 91.67%, the average recognition time of system is 3ms, running frequency of system is 57 MHz, and the utilize of FPGA chip resource is 74.6%, achieve the effects of easily recognition, small invasive, low cost, high recognition rate, rapid recognition speed and good stability.In summary, the speaker recognition system based on FPGA will hava board applicate prospects in social security, communication, banking, public security, national defense and other areas.