Voiceprint discrimination method and device, computer equipment and storage medium

Abstract

The invention relates to a voiceprint discrimination method and device, computer equipment and a storage medium. The method comprises the steps of obtaining a target correlation coefficient; based on the detected material data, performing preset processing on a voice sample in a detected material voice library so as to enable the voice quality of the voice sample in the detected material voice library to be matched with the voice quality of the detected material data; inputting the correlation coefficient of each comparison pair into a preset model, and outputting the log-likelihood ratio of each comparison pair; and according to the log-likelihood ratio of each pair, obtaining a preset threshold value, and based on the preset threshold value, discriminating the target correlation coefficient. The voice quality of the voice sample in the detected material voice library can be aligned with the voice quality of the detected material data, and the preset threshold value is fitted by the detected material voice library and the sample voice library, so that the preset threshold value can correspondingly change along with the change of the voice quality of the detected material data and is not a fixed value any more, then the subjectivity of voiceprint homology judgment can be greatly reduced, and the accuracy of voiceprint homology judgment is improved.

Description

technical field [0001] The present application relates to the technical field of voice recognition, in particular to a voiceprint discrimination method, device, computer equipment and storage medium. Background technique [0002] With the rapid development of voiceprint recognition technology, voiceprint comparison is now mainly based on deep learning methods. Specifically, the appraiser will input the inspection material data that needs to be compared into the deep learning system, and the system will compare the inspection material data with the numerous sample data collected in the system one by one, and the system will perform correlation analysis on each pair of data, And give the score of the similarity index (usually the distance evaluation of the voiceprint feature vector). The larger the score, the greater the possibility of homology between the two; otherwise, the greater the possibility of heterology. In an actual deep learning system, a threshold (usually an ex...

AI Technical Summary

Problems solved by technology

Since the external environment and channel may be different each time the data of the evidence is collected, a fixed threshold is always used to compare with the score of the similarity index, and misjudgment may occur because the threshold is set too small. Or misjudgment occurs due to excessive threshold setting, which seriously affects the accuracy of identification results

Images