Voice recognition system and voice recognition device

Abstract

The invention provides a voice recognition device and a system thereof, wherein the device comprises a noise reduction suppression module, which is used for carrying out noise suppression on user voice obtained in real time or other stored voice to obtain noise-reduced voice information; the noise reduction suppression module carries out noise suppression by adopting at least one of a spectrum removal method and/or a learning homologue method and/or a noise reduction automatic encoder; a conversion module, which is used for carrying out Laplace conversion on the sound information to obtain frequency spectrum information; a feature extraction module, which is used for obtaining a Meyer spectrum reciprocal and a linear cepstrum coefficient of the sound information to obtain a feature vectordescribing the sound information; a model training module, which is used for constructing a voice recognition model by using training samples, wherein the training of the voice recognition model utilizes the frequency spectrum information obtained by the conversion module and the characteristic vector extracted by the characteristic extraction module, inputting the characteristic vector obtained by the Meyer spectrum reciprocal and the linear cepstrum coefficient into an SVM layer; a sound recognition module, which is used for recognizing the sound information by using the trained sound recognition module.

Description

technical field [0001] Embodiments of the present invention relate to communication and computer technologies, and in particular to a voice recognition system and a voice recognition device. Background technique [0002] Voiceprint is the sound wave spectrum that carries speech information displayed by electroacoustic instruments. The production of human language is a complex physiological and physical process between the language center of the human body and the vocal organs. The vocal organs used by people when speaking, such as tongue and teeth , larynx, lungs, and nasal cavities vary greatly among individuals in size and shape, so any two people's voiceprints will be different. There are differences in the corresponding sound wave spectrum when different users speak, so the unique user can be identified through the voiceprint. [0003] In the prior art, the voiceprint recognition method has the defect of inaccurate recognition. Compared with face recognition, fingerprin...

AI Technical Summary

Problems solved by technology

[0008] After long-term experiments and theoretical research, the second penalty function was introduced. In the design process of multi-layer neural network, parameter adjustment has always been a difficult problem. Any improvement of parameters may lead to the improvement of recognition accuracy, so it is definitely not for those skilled in the art. What can be obtained without creative labor is not common knowledge

Images