Speech recognition method and device thereof and storage medium

Abstract

The invention discloses a voice recognition method and a device thereof and a storage medium, and the method comprises the steps: carrying out the coding of to-be-recognized voice data, and obtaining a first feature vector sequence; encoding each hot word in a preset hot word library to obtain a second feature vector sequence; encoding the audio clip of each hot word in the preset hot word library to obtain a third feature vector sequence; performing first attention operation on the first feature vector sequence and the third feature vector sequence to obtain a fourth feature vector sequence; and performing decoding operation according to the second feature vector sequence, the third feature vector sequence and the fourth feature vector sequence to obtain an identification result. By adopting the embodiment of the invention, the hot word recognition precision can be improved.

Description

technical field [0001] The present application relates to the technical field of speech recognition, in particular to a speech recognition method, device and storage medium. Background technique [0002] In the field of speech recognition, due to the low score of low-frequency words output by the end-to-end model, the effect of the traditional hot word score incentive method is limited. Google’s CLAS (Contextual Listen, Attend and Spell, CLAS) encourages hot words from the model level, and has achieved good results, but the method is too simple, and it is easy to misidentify sentences that do not contain hot words as hot words. Words lead to a decline in the overall recognition rate, and it is difficult to use them directly in the actual system. Therefore, the problem of how to improve the accuracy of hot word recognition needs to be solved urgently. Contents of the invention [0003] Embodiments of the present application provide a voice recognition method, device, and s...

AI Technical Summary

Problems solved by technology

[0002] In the field of speech recognition, due to the low score of low-frequency words output by the end-to-end model, the effect of the traditional hot word score incentive method is limited.

Google’s CLAS (Contextual Listen, Attend and Spell, CLAS) encourages hot words from the model level, and has achieved good results, but the method is too simple, and it is easy to misidentify sentences that do not contain hot words as hot words. words, leading to a decline in the overall recognition rate, and it is difficult to use them directly in the actual system. Therefore, the problem of how to improve the accuracy of hot word recognition needs to be solved urgently

Images