The Research And Application Of Voiceprint Recognition Technology

Voiceprint recognition technology as one of the three major characteristics of the biological recognition technology which is only second to the palmprint and fingerprint identification, is widely used in finance, justice, security and intelligence equipment. Compared with fingerprint and palmprint recognition technology, voiceprint recognition technology has the advantages of convenient collection, low cost and low algorithm complexity. Therefore, voiceprint recognition technology has a vast development space, and great commercial value and significance.Mel frequency cepstrum coefficient(MFCC) is the most common speech feature parameter for the voiceprint recognition, it is a kind of characteristic parameter which imitates the ear characteristic of the human ear. Compared with the linear spectrum, MFCC can well reflect the characteristics of the speech signal. This paper studies the extraction process of the MFCC and the wavelet packets decomposition relevant theoretical basis, aiming at the problem that the MFCC parameter extraction process ignored the high frequency detail and MFCC parameters can only reflect the static characteristics of the speech, this paper use the wavelet packet decomposition to improve the MFCC parameter extraction process. Experiments show that the improved feature parameter has a high recognition rate of voiceprint recognition and good anti noise.Support Vector Machine(SVM) is a machine learning method, It seeks to obtain optimal classification results in a finite number of samples. In this paper, we study the influence of SVM kernel function and its parameters on the classification of SVM, aiming at the selection of penalty factor and kernel parameter, use the particle swarm optimization(PSO) to optimize these two parameters. Then improved the PSO algorithm, improved the performance and the application of the method to the SVM. Experiments show that improved PSO algorithm to optimize the parameters of SVM, in the application of voiceprint recognition can achieve better classification effect, improve the recognition rate.In this paper, we mainly study the voiceprint recognition technology of speech signal processing, feature extraction and template matching algorithm, and the application of voiceprint recognition technology to the speaker verification system. Finally design a set of intelligent access control system based on speaker verification, and verifies the feasibility of the system is by experiments.