Speech recognition model training and speech recognition method and device

Abstract

The disclosure discloses a speech recognition model training and a speech recognition method, and relates to the technical fields of deep learning and speech processing. The training method of the speech recognition model comprises: obtaining training data; Constructing the neural network model comprising the first recognition model and the second recognition model; Inputting each speech data as the first input sequence into the first recognition model, according to the first recognition model for each The first output sequence and feature sequence of voice data output are obtained the second input sequence of each voice data; according to the second input sequence and label sequence of each voice data, the second recognition model is trained until the second recognition model converges, and the first A recognition model and a trained second recognition model are used as speech recognition models. The speech recognition method includes: acquiring speech data to be recognized; using the speech data to be recognized as an input of a speech recognition model, and using the output result of the speech recognition model as a recognition result of the speech data to be recognized.

Description

technical field [0001] The present disclosure relates to the technical field of data processing, in particular to the technical fields of deep learning and speech processing. Provided are a speech recognition model training and speech recognition method, device, electronic equipment and readable storage medium. Background technique [0002] Speech recognition is the conversion of sound signals into corresponding texts, which is one of the most important ways to realize human-computer interaction. In recent years, with the great improvement in speech recognition accuracy and the continuous popularization of smart devices, speech input has become one of the main ways of text input, and speech interaction has also been applied in more and more scenarios. The response speed and accuracy of speech recognition are key factors affecting the user experience of speech input and speech interaction. [0003] In terms of scenarios, speech recognition can be divided into streaming scen...

AI Technical Summary

Problems solved by technology

However, compared with non-streaming speech recognition, the accuracy of streaming recognition is relatively lower because it needs to start recognition before a sentence or a paragraph is finished.

Images