Font Size: a A A

The Audio Purification Technology Research Based On Low SNR

Posted on:2014-02-06Degree:MasterType:Thesis
Country:ChinaCandidate:B B DiFull Text:PDF
GTID:2248330395992292Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
The human voice is the quickest, most effective and most important way to exchangeinformation. With the rapid development of audio purification technology, The traditionalapproach has been unable to achieve the desired effect in low SNR environment, When thenoise seriously affect the performance of the system, so, how to improve the purificationeffect is necessary to be resolved at this stage. The purpose of this thesis is that from thesound signal with noise as possible to extract pure sound signal, and the background noisewill be suppression. So, the original voice can understand easy. The audio purificationtechnology research based on low SNR environment mainly used for listening area (catch thecriminals). Meanwhile, the technology involves a wide range of applications, Such as hearingaids, electrical cochlea, human-computer interaction systems and mobile voicecommunications and other aspects, they has high research value. In this context, the mainwork of this paper is as follows:(1) In this thesis, a zero rate on short-term average endpoint detection method beresearched. The method is separation using sound signals and noise signals, voice signals andaccurately determine the starting point of the noise signal, the completion of the preprocessingof the audio purification system;(2) Based on wavelet transform MFCC feature extraction method researched in thisthesis. This method represents an organic sound signal static characteristics and dynamicfeatures, MFCC can simulate human hearing characteristics, wavelet transform can be a gooddeal signals a change process, completed the sound signal to extract effective characteristicparameters;(3) This paper presents a neural network based on quantum hearing audio purificationmethod. This method implements the training process the sound signal pattern matching,using quantum neural network nonlinear mapping and self-learning ability to optimize the reduction parameters, thus completing the sound signal purification process. After a largenumber of experiments show that the method in the subjective and objective auditoryperformance indicators have shown significant improvement.
Keywords/Search Tags:Audio purification, low SNR, endpoint detection, feature extraction, quantum auditory neural network
PDF Full Text Request
Related items