Design And Implementation Of The Voice Access Control System Base On FPGA

With the rapid development of electronic technology, we more eager to exchange information between man and machine.Based on biometric identification technology is one of the important international research projects. On the current.Speech recognition as one of the biometric authentication technology, it according to the speech waveform which can reflect the speaker’s physiological parameters and behavioral parameters, automatically identify the speaker’s identity, therefore we pay more and more attention to the speech recognition. In recent years the rapid development of the special chip ASIC (Application Specific Integrated Circuit) which can process speech signal and the system SOC (System on Chip) which can recognize voice, create more favorable conditions for speech recognition, making speech recognition development more rapidly. In this paper, we will analyze the basic principles and system architecture of speech recognition, design and implementation of the voice access control system.In this paper, a systematic analysis of the status of the voice recognition inland and abroad, the purpose is improving the speech recognition system on the basis of speech recognition algorithms and processes. For real-time issues we put forward FPGA (Field Programmable Gate the Array) hardware platform to improve the solution of speech recognition algorithms, the system control module, information collection module, the algorithm module all have achieved on FPGA platform, give full play to the FPGA high-speed, parallel computing advantage. This article first introduces the concept of the speaker recognition system, and then analyzes the voice characteristic parameter extraction methods, as well as the speech recognition module. The system mainly consists five parts:the voice signal collection, endpoint detection, feature extraction, codebook training and recognition to determine.The voice collection module is controlled by WM8731voice codec chip and completed through the I2C bus; endpoint detection module was designed and implemented the dual-threshold algorithm based on short-time energy and short time zero crossing rate; has compared the both extraction algorithm of MFCC and LPCC, we choice the MFCC of feature extraction module; completed the training of the codebook by the LBG algorithm;the recognition algorithm by improved VQ.Experimental results show that the system has accomplished the text of voice access control system, the design of the entrance guard system can work normally speech, stable operation, strong ability to adapt and has the very high security. Through the improvement of the algorithm is effective to improve the system identification speed, is a real problem to solve the effective solutions.