Binaural speech separation method based on support vector machine

A technology of support vector machine and speech separation, applied in speech analysis, computer parts, instruments, etc., can solve the problems of loss of separated speech and audio points, unsatisfactory performance, etc.

Active Publication Date: 2018-05-29
SOUTHEAST UNIV
View PDF9 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

At present, the performance of commonly used binaural speech separation methods in complex acoustic environments is still unsatisfactory, and there is a phenomenon that the audio points of separated speech are lost

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Binaural speech separation method based on support vector machine
  • Binaural speech separation method based on support vector machine
  • Binaural speech separation method based on support vector machine

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0087] Such as figure 1 As shown, the support vector machine SVM speech separation method provided by the present embodiment includes the following steps:

[0088] Step 1: Convolve the training single-sound source speech signal with head-related impulse response function HRIR at different azimuths to generate multiple single-sound source binaural sound signals at different azimuths. Among them, the azimuth angle of the sound source is represented by θ, which defines that the front of the horizontal plane is 0°, and the range of θ is [-90°, 90°], with an interval of 10°, where -90° means the front left, and 90° means directly to the right;

[0089] Head-Related Impulse Response HRIR (Head-Related Impulse Response) is the time-domain representation of Head-Related Transfer Function (HRTF). The present invention adopts the HRTF database released by the Media Laboratory of Massachusetts Institute of Technology, which contains HRIR data of different elevation angles and different...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a binaural speech separation method based on a support vector machine. The method comprises the steps that after a binaural signal passes through a Gammatone filter, the interaural time difference ITD and the parameter interaural intensity difference IID of each sub-band acoustic signal are extracted; in a training phase, the sub-band ITD and IID parameters extracted from apure mixed binaural signal containing two sound sources are used as the input features of the support vector machine SVM, and the SVM classifier of each sub-band is trained; and in a test phase, in an environment with reverberation and noise, the sub-band features of a test mixed binaural signal containing two sound sources are extracted, and the SVM classifier of each sub-band is used to classify the feature parameters of each sub-band to separate each sound source in mixed speech. According to the invention, the method is based on the classification capability of the support vector machinemodel; robust binaural speech separation in a complex acoustic environment is realized; and the problem of frequency point data loss is effectively solved.

Description

technical field [0001] The invention relates to a speech separation method, in particular to a binaural speech separation method based on a support vector machine. Background technique [0002] Support Vector Machine (Support Vector Machine, SVM) is a binary classification model, which is a linear classifier with the largest interval defined in the feature space, and can achieve nonlinear classification by using different kernel functions. It shows many unique advantages in solving small sample, nonlinear and high-dimensional pattern recognition. At present, the performance of commonly used binaural speech separation methods in complex acoustic environments is still unsatisfactory, and there is a phenomenon of loss of separated speech audio points. Contents of the invention [0003] Purpose of the invention: the present invention aims at the problems existing in the prior art, and based on the high-dimensional and nonlinear classification capabilities of SVM, a binaural s...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G10L21/0308G06K9/62
CPCG10L21/0308H04S2420/01G06F18/2411
Inventor 周琳庄琰王立杰李楠
Owner SOUTHEAST UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap