The Study Of Speaker Recognition Based On Vector Quantization

Speakerrecognitionistheprocessof automaticallyrecognizingwhoisspeakingonthe basis of individual information include in speech signals. It has well applicationprospects in many fields. By analyzing the general principles and system structure ofspeaker recognition and considerating subsistent technology of speaker recognition,Linearpredictioncepstrumcoefficient(LPCC)andMelcepstrumcoefficient(MFCC)areadopted as characteristic parameters, the vector quantization (VQ) is used as speakerrecognition method to set up speaker recognition system. To improve the recognitioneffect,thetasksaremadeasfollows:1. Aim at the disadvantage of traditional double-gate threshold point detectionmethodthat cannot adapt the changeof environment,a extreme point detectionmethodbased on self-correlation function is brought up; it can detect and wipe off the silentsectionseffectivelyfromspeech.2. This article has modified the standard vector quantization (VQ) method andproposed a recognitionmethod that is Weight Distortion Measure VQbased on standarddeviation. It calculates the weight values that reflect the contribution of differentdimensionparameters,whichcanprovetherecognitioneffect.