Research On Systems For Voiceprint Recognition Based On Vector Quantization And Neural Network

With the development of computer science and technology, as well as the extensiveapplication of biological science in information science, the biometric authentication isthe current hot topics in the world of the computer. Voiceprint recognition technology asa kind of biometric authentication technology which has many advantages such as noteasy to lose, no need to remember and easy to use, meanwhile it also has some uniqueadvantages in the area of convenience, economy and accuracy that be compared withthe other way such as fingerprint, hand shape recognition, iris recognition. With theattention from the world, voiceprint recognition technology has become an importantand popular authentication mode in people’s daily life and work just because of thesecharacteristics.The research on voiceprint recognition based on vector quantization and neuralnetwork is deeply studied in this paper under the support of University Natural ScienceFoundation and major scientific and technological projects of Dazhou city. In this article,the author conducts many experiments on voiceprint recognition based on MATLABand samples of the voices of5speakers. The experimental results indicate that, thevoiceprint recognition system proposed in this paper can get better recognition effect.The major works of the thesis are as follows:First, study on the pretreatment technology, in the early stage of the voiceprintrecognition, including the solving of linear prediction coefficients (LPC), linearprediction cepstrum coefficient (LPCC) and Mel frequency cepstrum coefficient(MFCC) related in a general voiceprint recognition system, as well as the majortechnology in voiceprint recognition systems.Second, both BP and RBF Neural network are applied in voiceprint recognitionsystems, and the simulation experiments are conducted. Through the experimental data,the author analyzed the influence of various parameters on the system’s performance,and compared it with the traditional voiceprint recognition system based on distortion’smeasure.Third, the author analyzed the demand of individual oriented voiceprintrecognition, and studied the individual voiceprint recognition system based on the BPneural network by experimental verification and comparison.The structure of this paper is as follows: chapter1illustrates the meaning and objective of the project, and relevant the current research situation; chapter2introducesthe voiceprint recognition technology based theory, including voice preprocessing,feature extraction and main recognition methods; chapter3illustrates the research ofvoiceprint recognition system based on BP neural network, including relevanttheoretical analysis, system’s design and implementation as well as experimental resultsand analysis; chapter4illustrates the research of voiceprint recognition system based onRBF neural network, including relevant theoretical analysis, system’s design andimplementation as well as experimental results and analysis; chapter5illustrates avoiceprint recognition system surface to individual, including the system’s design andimplementation as well as experimental results and analysis; chapter six illustrates themajor contribution of the paper and the future work.