Research On The Voiceprint Identification Of Pitch Detection Algorithm

Speech recognition technology is the field of signal processing research at homeand abroad an important direction. In recent years, with the development of computertechnology and biometric technology driven, human speech recognition more andmore people’s attention, a lot of domestic and foreign experts and scholars havecarried out relevant studies.Speech recognition can be easily applied in real life. People do not have to worryabout forgotten passwords, identity and other issues can not be verified. Voiceprintidentification is a speech recognition, this identification can effectively determine theidentity of the speaker. For example, in judicial identification of the speaker’s identityconfirmation need to use voiceprint recognition technology. State to modify the new"Criminal Procedure" described, the voice can also be used as evidence, independentuse. This makes the voiceprint recognition can be used as a means of identificationapplied to a wide sphere.This paper introduces the wavelet decomposition and reconstruction tode-noising. As the adult female de-noising db6better, then continued on db6waveletimprovements to the adjustment coefficients of wavelet coefficients db6made achange to get a better deal with the effects of adult female can eventually achieve acertain frequency range similar Audio has a good de-noising effect. This article usesthe modified db6wavelet decomposition and reconstruction algorithm6layers.Experimental results show that to achieve the desired results.Then voiceprint identification is proposed to detect several commonly usedtraditional methods of auto-correlation function (Autocorrelation Function, ACF), theaverage magnitude difference function (Average Magnitude Different Function,AMDF), variable length, the average magnitude difference function (Length VariedAverage Magnitude Difference Function, LV-AMDF) and some traditional classicalalgorithm, due to the above algorithm has some unavoidable disadvantages, so thisarticle to improve it. Application ACF combined with the LV-AMDF improved algorithm, this algorithm has good noise immunity, the extraction can be stabilizedpitch. The improved algorithm inherits the advantages of traditional algorithms, whileits noise impact over the previous several classical algorithms significantly reduced,but also overcome certain formants, frequency, second harmonic in judging errors,improves the noisy circumstances where the accuracy of pitch, pitch detectionimproves the robustness and accuracy. Experimental test, after the improvedalgorithm effectively extract the adult female of the pitch period, reducing the falsepositive rate, making Voiceprint recognition rate has been improved. This algorithmis simple, inherits the advantages of traditional algorithms, but also overcomes someof the shortcomings of other algorithms can be more accurate pitch.