Language recognition feature fusion method used in low signal-to-noise ratio environment

Abstract

The invention discloses a feature fusion method for language recognition in a low signal-to-noise ratio environment, and belongs to the field of voice recognition. The problems that in the prior art,engineering application is difficult, and the recognition rate is low under a low signal-to-noise ratio are solved. Language identification, extraction of effective features and reduction of noise influence are the key of accurate identification. According to the method, syllable segmentation, CFCC coefficients, principal component analysis and Teager energy operator cepstrum parameters are mainlyadopted. According to the invention, syllable segmentation is carried out on a full voice segment, and then a CFCC coefficient is extracted from each syllable voice segment; principal component analysis is carried out on the extracted CFCC coefficients by using a PCA technology, and selecting a first S frame with the highest contribution rate from an F frame corresponding to each syllable; For improving the robustness of features, Teager energy operator cepstrum parameters based on syllable extraction are fused to obtain a fused feature set. the extracted fusion feature set is inputted into alanguage recognition model to train a corresponding language recognition model, the trained language model is mounted to a server side, to-be-recognized voice is acquired through a client and is inputted into the server, the fusion features are extracted, and scoring judgment is carried out on the fusion features and the trained language model; and finally an identification result is outputted and returned to the client. Tests show that the method can improve the accuracy of language recognition in a low signal-to-noise ratio environment, and is high in operation speed and small in calculatedamount.

Description

technical field [0001] The invention relates to a feature fusion method for language recognition in a low signal-to-noise ratio environment, belonging to the field of speech recognition. Background technique [0002] With the steady progress of globalization and a community with a shared future for mankind, the issue of language recognition has gradually gained attention. The problem of communication between people of different countries is a major obstacle to the development of globalization. People are no longer satisfied with information interaction with smart devices only through keyboards and monitors, but urgently need a more natural and more accessible to most people. The accepted way is to communicate with smart devices, so that computers can understand human speech, or use voice to control various smart devices. Using the most direct and convenient language for human to exchange information to communicate with computers has always been a research topic that has att...

AI Technical Summary

Problems solved by technology

The accuracy of current language recognition technology for noise-free language recognition is good enough, but the recognition rate of languages ​​with low signal-to-noise ratio needs to be improved

Images