A method and device for cross-language speech recognition

Abstract

The invention discloses a method and device for cross-language speech recognition, and relates to the technical field of speech processing. A specific implementation of the method includes: obtaining cross-lingual sample data, using the sample data as the input data of the preset neural network model for training to obtain a language class discriminator; inputting the audio to be identified into the language class The discriminator divides the audio to be recognized according to the language type determined by the language type discriminator; uses the recognition engine corresponding to the determined language type to recognize the segmented audio to be recognized respectively. This embodiment does not need to modify the existing speech recognition engine, and has low cost, high recognition rate and high accuracy.

Description

technical field [0001] The invention relates to the field of speech processing, in particular to a cross-language speech recognition method and device. Background technique [0002] The intelligence and integration of electronic equipment are getting higher and higher, and the traditional information retrieval and menu operation methods are increasingly unable to meet the requirements. There is an urgent need for a more convenient information retrieval and command operation method to replace the traditional button operation. Speech recognition technology came into being. However, in most traditional automatic speech recognition systems, only the most commonly used language of the country is supported, and other languages ​​are less supported or not supported. For this situation, the conventional approach is: (1) Different languages ​​are considered independently, and a language model is trained from scratch for each language. (2) In the acoustic model, the most commonly us...

AI Technical Summary

Problems solved by technology

[0003] In the process of realizing the present invention, the inventors found that there are at least the following problems in the prior art: (1) training a language model for a language from scratch requires a large amount of manually labeled data, which is not only expensive, but also takes a lot of time to obtain; building separate language models for each language hinders smooth recognition and increases the cost of recognizing mixed-language speech

(2) The training takes a long time, and the input of manpower and material resources is large. The range of other language recognition is based on the field of investment. The overall support range is relatively narrow, and there are similar pronunciations, which may easily cause misidentification and low accuracy.

Images