Design And Implementation Of Speaker Recognition System Based On SOPC

Speaker recognition refers to the speaker through the analysis of speech signal processing to automatically identify the speaker’s identity, which have a good prospect in many areas.As the speaker recognition technology continues to develop,In addition to the reliability of recognition algorithms,People have higher expectations on identification of equipment about the miniaturization、portable and scalable.Therefore, implementation of speaker recognition has great practical value in embedded systems.This article propose a design about embedded soft-core processor speaker recognition system based on SOPC and Nios Ⅱ.Cyclone FPGA chip as the core of the system, Design approach use combination with hardware and software.Application of Altera’s Nios Ⅱ IDE and Quartus Ⅱ software to design and implement speaker recognition system based on SOPC(System on Programmable Chip) in the Cyclone Ⅱ family EP2C35F672C6chips. Hardware use Quartus Ⅱ software to design and implement,including voice data collection、 pre-emphasis、FFT(Fast Fourier Transform) and hardware acceleration etc.The part of software is designed by the Nios Ⅱ soft-core to implement endpoint detection、feature extraction training model、identify matching、 keyboard control and LCD display etc. The characteristics extracted from the audio signal in system is MFCC(Mel-Frequency Cepstral coefficient),using vector quantization (VQ) in the recognition algorithm. All algorithms are written by C code to implement in the Nios Ⅱ IDE.Test showed that the function of hardware is effectively、feasible, software and hardware work together to achieve data collection, storage and processing on FPGA. The system as embedded systems is easy to program, debug and easy to use. This system has a very high cost-effective, can saving software and hardware resources, has good application prospects.