A speech recognition method and system

Abstract

The invention discloses a voice recognition method and a system thereof. The method comprises the following steps of receiving a voice signal to be recognized and extracting an acoustic characteristic of the voice signal; according to the acoustic characteristic and an acoustic model which is trained in advance, acquiring a posterior probability of each modeling unit; carrying out compression smoothness processing on the posterior probability of the modeling unit and acquiring a posterior probability processed by each modeling unit; and decoding the posterior probabilities of all the frames of the voice signal and acquiring a voice recognition result, wherein the posterior probabilities are processed by the modeling unit. In the invention, after the posterior probability of each modeling unit is acquired, the compression smoothness processing is performed on distribution so that a coverage rate of a plurality of candidate sequences corresponding to the posterior probability of each modeling unit to a correct voice recognition result is increased and a voice recognition effect is further improved.

Description

technical field [0001] The invention relates to the field of voice recognition, in particular to a voice recognition method and system. Background technique [0002] With the increasing development of Internet technology, speech recognition technology has entered various fields such as industry, communication, home appliances, automobiles, home services, and medicine. Speech recognition technology is a technology that allows machines to convert voice signals into corresponding text or commands through the process of recognition and understanding. Speech recognition technology mainly includes three aspects: feature extraction technology, pattern matching criteria and model training technology. [0003] With the rapid development of neural networks, such as deep neural network DNN technology, speech recognition systems based on neural networks have become the most mainstream speech recognition technology. In the training process of the neural network used for speech recognit...

AI Technical Summary

Problems solved by technology

[0004] The embodiment of the present invention provides a speech recognition method and system to solve the problem that when the training data is small, the speech recognition system based on the neural network often causes errors to be amplified for pronunciations not covered by the training data, so as to improve the recognition of continuous speech Effect

Images