Speech recognition method, device, electronic equipment, and readable storage medium

Abstract

The present invention provides a voice recognition method, device, electronic equipment, and readable storage medium, including: inputting audio information of a user to be identified into a trained real-time voice recognition acoustic model to obtain a recognition result of the audio information of the user to be identified; wherein , the real-time speech recognition acoustic model is obtained by training according to user audio training samples carrying audio text labels; wherein, the user audio training samples are obtained by combining preset real-time speech recognition system and preset non-real-time recognition system It is obtained by unsupervised screening of the audio recognition results of the same user. Through the method of the present invention, based on the recognition results of the real-time speech recognition system and the recognition results of the non-real-time recognition system, the training data is automatically extracted and screened in an unsupervised manner to iteratively train the real-time speech recognition acoustic model, which can save manpower and time , to improve the voice recognition effect and enhance the user experience.

Description

technical field [0001] The invention relates to the field of voice recognition, in particular to a voice recognition method, device, electronic equipment, and a readable storage medium. Background technique [0002] With the increasing penetration rate of artificial intelligence products in life, speech recognition, as a basic way of natural human-computer interaction, has a wide range of applications in smart homes, smart cars, smart customer service and other scenarios. As one of the prerequisites for natural human-computer interaction, speech recognition is bound to be the development trend of artificial intelligence. Only when the accuracy of speech recognition is close to perfect can natural human-computer interaction continue to develop. [0003] In the application scenario of real-time recognition of voice robots, training data is currently obtained by marking a large amount of customer online data, and then the acoustic model is updated. Since labeling data requires...

AI Technical Summary

Problems solved by technology

Since labeling data requires a lot of manpower and takes a long period of time, and the quality inspection of the data is required after the data labeling is completed, it takes too long for the recognition effect to be updated online, and the user experience is poor.

Images