Voice quality evaluation method and device

Abstract

The invention provides a voice quality evaluation method and device, and relates to the field of artificial intelligence. The method comprises the following steps: enhancing a voice signal acquired by equipment through a voice enhancement algorithm to obtain enhanced voice; the enhanced voice is input to a voice quality evaluation model, loss information of the enhanced voice is obtained, the voice quality evaluation model comprises at least two acoustic models, and the loss information is determined based on the fact that decoding results of decoders of the at least two acoustic models on the enhanced voice serve as true values; and judging whether the quality of the enhanced voice reaches the standard or not based on the loss information. The voice quality evaluation method and device can be decoupled from a specific voice enhancement algorithm and a specific ASR rear end, can reflect the influence of the voice enhancement algorithm on the ASR rear end, and can be flexibly applied to various different scenes.

Description

technical field [0001] The present application relates to the field of artificial intelligence, in particular to a voice quality evaluation method and device. Background technique [0002] With the development of artificial intelligence technology, automatic speech recognition (automatic speech recognition, ASR) technology is becoming more and more important. ASR is the key technology to realize human-computer interaction. Machines can recognize and understand human speech through ASR technology, convert speech into text, or convert speech into commands that the machine can understand, and perform corresponding operations according to the commands. [0003] Due to the existence of noise and interference in the actual environment, it is necessary to perform speech enhancement processing on the collected speech signals before performing speech recognition. Speech enhancement is to use a speech enhancement algorithm to extract the original speech as pure as possible from the n...

AI Technical Summary

Problems solved by technology

[0004] The speech quality evaluation models used in the current speech quality evaluation methods are all based on the processed speech signal, which has nothing to do with the specific ASR backend, and cannot show the impact of the change of the speech enhancement algorithm on ASR. For example, when ASR changes from a When one device is extended to another device, the speech enhancement algorithm changes, and the enhanced speech also changes. The index obtained by the original speech quality evaluation algorithm cannot reflect the impact of the speech enhancement algorithm change on the ASR backend.

Images