Depth and breadth neural network combined speech enhancement algorithm of digital hearing aid

A neural network and speech enhancement technology, applied in hearing aids, hearing aid signal processing, electrical components, etc., can solve problems such as increased algorithm complexity, achieve strong robustness, improve intelligibility, and improve voice quality

Active Publication Date: 2016-05-25
BEIJING UNIV OF TECH
View PDF3 Cites 27 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Obviously, there is no speech enhancement algorithm that can adapt to all noise environments, which requires classifying the noise, and using different speech en

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
  • Depth and breadth neural network combined speech enhancement algorithm of digital hearing aid
  • Depth and breadth neural network combined speech enhancement algorithm of digital hearing aid
  • Depth and breadth neural network combined speech enhancement algorithm of digital hearing aid

Examples

Experimental program
Comparison scheme
Effect test

Example Embodiment

[0024] Step 1: Pre-processing the input signal of the digital hearing aid, that is, de-averaging, normalization, pre-emphasis, framing, and windowing;

[0025] (1) The analog input signal x(t) of the digital hearing aid, during the A / D conversion, the sampling rate is 16kHz, and the converted digital signal is x(n);

[0026] (2) In order to eliminate the DC component, the average is removed; in order to cancel the magnitude difference between the data of each dimension, and avoid the large error caused by the large magnitude difference between the input and output data, the data is normalized; due to the high frequency band energy of the voice signal Small, resulting in weakening of high-frequency transmission, and pre-emphasis emphasizes high-frequency parts to improve signal transmission quality. The de-averaging formula is shown in formula (1), the normalization formula is shown in formula (2), and the pre-emphasis formula is shown in formula (3).

[0027] x 1 (n)=x(n)-mean(x(n))...

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 depth and breadth neural network combined speech enhancement algorithm of a digital hearing aid belonging to the speech signal processing technical field. The method comprises following steps: firstly, carrying out speech activity detection to noise contained speech signals; extracting features such as autocorrelation function maximum values and variances of frequency bands of the noise contained speech signals; building a two-value decision device by using a BP neural network; judging speech segments and noise segments; secondly, extracting MFCC and first level MFCC features; detecting noise types by a depth neural network, wherein the depth neural network is formed by cascading a learning vector quantization neural network and the BP neural network; finally, building a breadth neural network formed by connecting various networks in parallel; automatically selecting corresponding neural networks by the breadth neural network according to the noise types; removing the noises, thus obtaining the enhanced speech and improving the speech intelligibility of the hearing aid output speech. According to the algorithm, the training processes of the neural networks are finished offline; the complexity of the test algorithm of the trained networks is low; and therefore, the real timeliness is satisfied.

Description

technical field [0001] The invention belongs to the technical field of speech signal processing and relates to two key speech signal processing technologies of noise classification and speech enhancement in digital hearing aids. Background technique [0002] The transmission of information through voice is the most important, most effective, most commonly used and most convenient form of information exchange for human beings. However, for deaf and hearing-impaired patients, the inability to carry out normal voice communication will seriously affect their daily life. With the aging of the global social population becoming more and more obvious, the number of deaf and hearing-impaired patients is also increasing day by day, and the problem of hearing impairment is becoming more and more prominent. However, for most patients, wearing a suitable digital hearing aid can effectively improve their hearing status and greatly improve their hearing level. At present, there are many ...

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
IPC IPC(8): H04R25/00
CPCH04R25/507H04R2225/43
Inventor 李如玮时勇强
Owner BEIJING UNIV OF TECH
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