Speech endpoint detection method and device based on statistical model

Abstract

The present invention proposes a voice endpoint detection method and device based on a statistical model, wherein the method includes: receiving an input voice signal to be detected; extracting the first voice feature information of the voice signal to be detected by frame, and analyzing the first voice feature The information is subjected to anti-noise processing to generate the second speech feature information of the speech signal to be detected; the recognition result of the speech signal to be detected is generated according to the second speech feature information and the acoustic model; the speech signal to be detected is initially detected according to the recognition result and the preset silence detection algorithm. detecting the speech endpoint of the speech signal; and calculating confidence information of the speech signal to be detected, and adjusting the speech endpoint according to the confidence information. The speech endpoint detection method based on the statistical model in the embodiment of the present invention accurately locates the speech endpoint of the speech signal to be detected, improves the accuracy of speech endpoint detection, and can further improve the accuracy of speech recognition and improve the performance of speech recognition.

Description

technical field [0001] The invention relates to the technical field of speech recognition, in particular to a method and device for detecting speech endpoints based on statistical models. Background technique [0002] With the development of human-computer information interaction technology, speech recognition technology shows its importance. In speech recognition system, speech endpoint detection is one of the key technologies in speech recognition. Speech endpoint detection refers to finding the starting point and ending point of the speech part in the continuous sound signal. Whether the endpoint detection is accurate or not will directly affect the performance of the speech recognition system. If there is an error in endpoint segmentation, it will lead to missed recognition or misrecognition, which in turn can lead to inaccurate speech recognition results. [0003] At present, the traditional speech endpoint detection method is mainly to obtain the energy in time doma...

AI Technical Summary

Problems solved by technology

[0004] However, in the process of realizing the present invention, the inventors have found that the above-mentioned speech endpoint detection algorithm has at least the following problems: (1) the above-mentioned speech endpoint detection method is suitable for stationary noise and high SNR environment, but it is not suitable for non-stationary noise and relatively high signal-to-noise ratio. In a low SNR environment, the detection effect of the above speech endpoint detection method is not good, and the accuracy of the detected speech endpoint is low; (2) for speech signals under different SNRs, it is difficult to select a suitable threshold, and it is impossible to Ensure detection accuracy in quiet environment and detection accuracy in noisy environment

Images