A voice recognition method and device

Abstract

The invention discloses a speech recognition method and device. The method includes the following steps that: speech to be recognized is determined; acoustic feature extraction is performed on the speech to be recognized; extracted acoustic features are inputted into a decoder; the decoder is adopted to call a trained language model and a trained acoustic model so as to recognize the extracted acoustic features, so that a recognition result can be obtained, a deep neural network-Hidden Markov model (DNN-HMM) is adopted as the acoustic model, and two parallel tasks, namely speaker identity confirmation and phoneme posterior probability learning, are added to the output layer of the DNN-HMM during the training process of the DNN-HMM; and the recognition result is outputted through the decoder.

Description

technical field [0001] The invention relates to electronic technology, in particular to a voice recognition method and device. Background technique [0002] Speech recognition technology is a technology that converts speech signals into text symbols recognizable by computers, and solves the problem of allowing machines to understand human speech. Speech recognition technology has an English recognition accuracy rate of about 80% in complex environments such as telephone calls, conferences, and customer service, which is still far from the human error rate of 2% to 4%. In fact, the current speech recognition technology has defects in the following aspects: 1) recognition and understanding of natural language. First, continuous speech must be decomposed into units such as words and phonemes, and secondly, a rule for understanding semantics must be established. In fact, human voices are vast, and the semantics are so wonderful that some normal communication between people has...

AI Technical Summary

Problems solved by technology

In fact, the current speech recognition technology has defects in the following aspects: 1) recognition and understanding of natural language

2) Large amount of voice information

5) Environmental noise and interference have a serious impact on speech recognition, resulting in low recognition rate

6) Audio processing technology is limited, the human ear can clearly distinguish two pronunciations, and its acoustic characteristics may be almost the same for the speech recognition engine

[0003] Due to the above possible shortcomings of the current speech recognition technology, the accuracy of speech recognition is not high and cannot meet the accuracy requirements of users; although the accuracy of some speech recognition can meet people's requirements for accuracy, speech recognition The parameters and the complexity of the model will be greatly increased, resulting in an increase in product cost, and users will need to pay a lot of extra fees to improve the accuracy of speech recognition

Images