Voiceprint recognition method based on pitch period mixed characteristic parameters

Abstract

The invention provides a voiceprint recognition method based on pitch period mixed characteristic parameters. The method comprises the following steps of voice signal acquisition and input, voice signal preprocessing and voice signal combined characteristic parameter extraction, i.e. a pitch period, LPCC, delta LPCC, energy, first order difference of energy and GFCC characteristic parameters are extracted to be combined into multidimensional characteristic vectors together, the multidimensional characteristic vectors are screened by adopting a discrete binary particle swarm optimization algorithm, the voice model of a speaker is obtained by introducing universal background model UBM training, and finally test voice is recognized by utilizing a GMM-UBM model. Compared with a mode that voiceprint recognition is performed through single voice signal characteristic parameter, recognition accuracy of the voiceprint recognition and system stability are effectively enhanced by adopting the combined characteristic parameters and using the voiceprint recognition system of the GMM-UBM model.

Description

technical field [0001] The invention relates to the field of speech signal processing, in particular to a voiceprint recognition method based on pitch period mixing characteristic parameters. Background technique [0002] Under the premise of today's information age, identification technology, one of the important components of information security, has brought new challenges. Due to the limitations of the algorithm and the rise of hardware and software decryption technology, the traditional password identification has shown its disadvantages. As one of the new technologies of identification, voiceprint recognition technology has attracted more and more attention because of its unique advantages of convenience, economy and accuracy. [0003] Voiceprint recognition is to extract the speaker's personality characteristics from a segment of the speaker's voice, and through the analysis and identification of personal characteristics, so as to achieve the purpose of identifying o...

AI Technical Summary

Problems solved by technology

[0007] The present application provides a voiceprint recognition method based on pitch cycle mixed feature parameters, which includes the following steps: voice signal acquisition and input, voice signal preprocessing, and voice signal combination feature parameter extraction: that is, extracting pitch cycle, LPCC, △LPCC, Energy, the first-order difference of energy, and GFCC feature parameters are combined into a multidimensional feature vector, and the discrete binary particle swarm optimization algorithm BPSO is used to screen the combined feature parameters of the speech signal, and the general background model UBM is introduced to train the speaker's voice model. The GMM-UBM model recognizes the test voice to solve the technical problems of low recognition accuracy and unstable voiceprint recognition system using a single voice parameter in the prior art

Images