Chinese speech recognition method based on cyclic neural network language model and deep neural network acoustic model

Abstract

The invention discloses a Chinese speech recognition method based on a cyclic neural network language model and a deep neural network acoustic model. The Chinese speech recognition method mainly includes the following steps: S1, training the language model based on a cyclic neural network; S2, training the acoustic model based on a deep neural network; and S3, employing a Viterbi search scheme bya decoder for Chinese speech recognition based on the cyclic neural network language model and the deep neural network acoustic model. The Chinese speech recognition method based on a cyclic neural network language model and a deep neural network acoustic model combines the accuracy of the cyclic neural network and the low delay of the deep neural network, solves the shortcomings that a current n-gram language model has low accuracy and a long and short time memory acoustic model has high time delay, and achieves Chinese speech recognition with low delay and higher accuracy.

Description

technical field [0001] The invention relates to speech recognition technology and deep learning technology, in particular to a speech recognition method based on a recurrent neural network language model and a deep neural network acoustic model. Background technique [0002] With the popularity of smart hardware products such as Amazon Echo smart speakers, as the most important means of human-computer interaction, the market for voice recognition is also rising. According to the "2015-2020 Global and China Voice Industry Report" released by the internationally renowned market research company Research and Markets in 2016, as the application of voice in the smart industry continues to deepen, by 2020, the global voice market is expected to reach 19.17 billion US dollars . [0003] The mainstream representative of traditional continuous speech recognition technology is GMM-HMM. Around 2011, Microsoft, Google and other companies began to apply deep neural networks to speech re...

AI Technical Summary

Problems solved by technology

[0014] The main purpose of the present invention is to overcome the shortcoming and deficiency of prior art, provide a kind of speech recognition method based on recurrent neural network language model and deep neural network acoustic model, combine the accuracy of recurrent neural network and the low delay of deep neural network , which solves the shortcomings of the low accuracy of the existing n-gram language model and the high latency of the long-short-term memory network (LSTM) acoustic model

Images