Research On The Access Control System Based On Speaker Recognition

Evolved over the years, security industry in China about access control has formed anumber of advanced industrial systems. Integration of biometrics access control system hasbecome a trend, which is reflected in a phenomenon, that people use different voices, differentfingerprints, different iris, and other personalized information to complete the identification inthe access control, and the number of such systems has been growing rapidly. In this article,remote authenticated speaker recognition technology has been used in intelligent accesscontrol system. It has achieved a natural communication between specific person and machine,and it has a certain degree of security and convenience.Background with the development status of access control system, it proposed that usingspeech recognition technology in the DSP chip to complete the design of access controlsystem. Firstly, it introduced the function can be achieved and the classification of speakerrecognition technology, then it made a detailed analysis of the basic theory in speakerrecognition technology. In the pre-processing, using Sampling, quantifying, pre-emphasis,framing, windowing to process the speech signal, and it used Matlab to simulate the aboveprocesses; it also used double threshold endpoint detection method to detect the start and theend of the voice; In the feature extraction, it had a detailed analysis of Linear PredictiveCepstral Coding (LPCC) and Mel Frequency Cepstrum Coefficient (MFCC), and selectingMFCC as a speaker recognition feature parameter for this system. About identificationmethods, firstly it described the classic Dynamic Time Warping (DTW) algorithm, which canidentify small vocabulary with good results, and then it proposed Genetic Algorithm DynamicTime Warping (GA-DTW) to break through. Combined with the characteristics of DTW, alsohas global search and parallel computing capabilities. By the results of the simulation inMatlab for, it has been proved that the new algorithm can have more rapid and accurateidentification for isolated words.It is an embedded speaker recognition system based on TMS320C6713DSP chip. Itintroduced the structure of TMS320C6713and TLV320AIC23, and had a detailed descriptionof the seamless connection between the two chips. It also designed external expansionmemory, power module, reset circuit, and JTAG circuit. In the CCS development environment,improved C program let each module be properly initialized, and the system can achieve the bootstrap operation. After the system testing, the experimental results and performanceanalysis show that real-time speaker recognition access control designed in this paper can runproperly.