Voice detection method and device, prediction model training method and device, equipment and medium

Abstract

The invention discloses a voice detection method and device, a prediction model training method and device, equipment and a medium, and belongs to the technical field of voice interaction. The invention relates to a multi-mode voice end point detection method, which comprises the following steps: identifying a shot face image through a model so as to predict whether a user has an intention of continuing speaking or not, and judging whether a collected audio signal is a voice end point or not by combining a prediction result. In addition, the characteristics of the visual mode of the face imageare fused for detection on the basis of acoustic characteristics, and whether the voice signal is a voice end point or not can be accurately judged by utilizing the face image even if the backgroundnoise is very strong or the user pauses during speaking, so that the interference caused by the background noise and speaking pauses is avoided. Therefore, the problem that the voice interaction is detected to be in an end state too late or too early due to interference of background noise and speaking pause is avoided, and the accuracy of detecting the voice end point is improved.

Description

technical field [0001] The present application relates to the field of voice interaction technology, and in particular to a voice detection method, a prediction model training method, a device, a device, and a medium. Background technique [0002] In the voice interaction technology, in order to realize the human-computer interaction function based on voice, the voice start point and the voice end point in a piece of voice are usually recognized, and the part between the voice start point and the voice end point is intercepted as a voice command. Instructions to instruct the device to perform corresponding operations. Among them, the voice start point is usually triggered by the user's active operation, which is easily determined by the collection time point of the wake-up word, the time point when the voice interaction activation option is triggered, and the voice end point needs to be determined by the device through voice analysis. processing can be obtained. It can be ...

AI Technical Summary

Problems solved by technology

[0004] When the above method is used to detect the end point of the speech, once the background noise is strong, the tail silence duration of the detected audio signal will be longer than the accurate tail silence duration, resulting in the speech end point being easily missed, which will lead to too late Detect that the voice interaction has ended; in addition, once the user pauses during speaking, the tail silence duration of the detected audio signal is smaller than the accurate tail silence duration, which will cause the voice interaction to be detected prematurely in final state

It can be seen that the accuracy of the speech end point detected by this method is poor

Images