Method and apparatus for extracting voiceprint spectrum

Abstract

The invention provides a method and apparatus for extracting a voiceprint spectrum. The method includes: acquiring a sample voice signal and converting the same into a log-domain spectrum; extracting the voiceprint factor of each frame of the sample voice signal by using a pre-trained voiceprint factor extraction model and extracting the content factor of each frame of the sample voice signal by using a pre-trained content factor extraction model based on the log-domain spectrum; training a content spectrum reconstruction model for reconstructing the content factor into a content spectrum and a voiceprint spectrum reconstruction model for reconstructing the voiceprint factor into a voiceprint spectrum based on the extracted content factor and the voiceprint factor; and discarding the content spectrum reconstruction model, inputting a target voice signal into the voiceprint spectrum reconstruction model to obtain the voiceprint spectrum in the target voice signal. The method and apparatus can accurately extract the voiceprint spectrum from the voice signal so as to subsequently and clearly observe the characteristics of a speaker according to the extracted voiceprint spectrum to help improvement in the accuracy of the voiceprint verification in the fields of forensic identification and identity identification.

Description

technical field [0001] The invention relates to the technical field of speech signal processing, in particular to a method and device for extracting voiceprint spectrum. Background technique [0002] Voiceprint is to judge the characteristics of the speaker contained in the speech signal. In judicial practice, voiceprint is one of the effective tools to verify the target, and it is of great significance in the fields of forensic identification and identification. [0003] The traditional voiceprint comparison method is generally realized through the spectrum, that is, the sound (that is, the voice signal) is converted into a spectrum first, and then the identification expert judges whether the two voices come from the same speaker by observing the spectrum of the two voices. . [0004] However, the spectrum of these two voices contains not only speaker information, but also speech content information. Usually, the change of speech content is more significant, which means t...

AI Technical Summary

Problems solved by technology

[0004] However, the spectrum of these two voices contains not only speaker information, but also speech content information. Usually, the change of speech content is more significant, which means that most of the information seen by identification experts through the spectrum is related to pronunciation

Images