Voice recognition method and device and computer readable storage medium

Abstract

The invention relates to the technical field of voice processing, particularly provides a voice recognition method and device and a computer readable storage medium, and aims to solve the technical problem of how to accurately and efficiently perform voice recognition. Therefore, according to the method provided by the embodiment of the invention, a knowledge distillation algorithm can be adopted to enable a plurality of trained first speech recognition models to guide a second speech recognition model to carry out model training; the phoneme recognition capability and decoding capability of the trained second speech recognition model on the input speech are close to the phoneme recognition and decoding capabilities of the plurality of first speech recognition models; therefore, only one second voice recognition model needs to be operated on the computer equipment, the voice recognition effect of parallel operation of the multiple first voice recognition models can be achieved, and the voice recognition efficiency and accuracy are remarkably improved.

Description

technical field [0001] The present invention relates to the technical field of speech processing, in particular to a speech recognition method, device and computer-readable storage medium. Background technique [0002] Speech recognition refers to the semantic analysis of speech signals to obtain the text information contained in the speech signals, such as converting speech signals into Chinese text information. The current conventional speech recognition method mainly uses training samples to train the speech recognition model, so that the trained speech recognition model has the ability of speech recognition, and then can use the trained speech recognition model to perform speech recognition on the speech to be recognized. At present, in addition to using a single speech recognition model for speech recognition, multiple speech recognition models can also be used for speech recognition at the same time, and then comprehensively analyze each speech recognition result to de...

AI Technical Summary

Problems solved by technology

However, in practical applications, due to the limitations of equipment configuration and cost, computer equipment equipped with speech recognition models often cannot run multiple speech recognition models in parallel.

If you use serial processing to run each speech recognition model in turn, and then conduct comprehensive analysis to determine the final recognition result based on each speech recognition result, it will greatly affect the speed of speech recognition

Images