Voice recognition method and device based on Chinese and English mixed dictionary

Abstract

The invention proposes a voice recognition method and device based on a Chinese and English mixed dictionary, and the method comprises the steps: obtaining the Chinese and English mixed dictionary marked with the IPA (International Phonetic Alphabet), wherein the Chinese and English mixed dictionary comprises a Chinese dictionary and an English dictionary which passes through the Chinglish correction; taking the Chinese and English mixed dictionary as a training dictionary, adding one CNN (Convolutional Neural Network) to five time recursion neural network LSTMs to form a model, taking the syllables or words as the targets, taking a CTC (Connectionist Temporal Classifier) as a training rule for the training of the model, and obtaining a trained CTC acoustic model; combining with the trained CTC acoustic model to carry out the voice recognition of the Chinese and English mixed language. According to the embodiment of the invention, the method employs the Chinese and English mixed dictionary for training, wherein the Chinese and English mixed dictionary comprises the Chinese dictionary and the English dictionary which passes through the Chinglish correction. The Chinese and English words are complete, and the method and device can achieve the recognition of Chinglish. The method and device further improve the recognition accuracy of the Chinese and English mixed language through combining with the CTC acoustic model.

Description

technical field [0001] The invention relates to the technical field of voice recognition, in particular to a voice recognition method and device based on a Chinese-English mixed dictionary. Background technique [0002] Currently, with the globalization of life, the phenomenon of using mixed languages ​​to communicate has become a common phenomenon. Statistically, there are more people who speak multiple languages ​​than monolingual speakers. Acoustics between mixed languages ​​and complexities between languages ​​pose challenges for speech recognition. Therefore, the study of mixed language acoustic models is an important research direction. [0003] Hybrid speech recognition technology refers to the use of Chinese and English mixed dictionaries to train the mixed language acoustic model to obtain a speech recognition model. At present, the method of obtaining a Chinese-English mixed dictionary is to obtain a Chinese dictionary including a phoneme set marked with consona...

AI Technical Summary

Problems solved by technology

[0005] For this reason, the first object of the present invention is to propose a kind of speech recognition method based on Chinese-English mixed dictionary, for solving the low problem of speech recognition accuracy in the prior art

Images