Convolutional neural network (CNN)-based sound source positioning algorithm

A convolutional neural network and sound source localization technology, applied in the information field, can solve the problems of sound source localization noise and reverberation, and achieve the effect of convenient and simple data collection and good localization effect.

Active Publication Date: 2018-02-16
NANJING UNIV OF POSTS & TELECOMM
View PDF2 Cites 28 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The implementation of this algorithm effectively solves the problems of noise and reverberation in traditional sound source localization

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
  • Convolutional neural network (CNN)-based sound source positioning algorithm
  • Convolutional neural network (CNN)-based sound source positioning algorithm
  • Convolutional neural network (CNN)-based sound source positioning algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0053] The invention will be described in further detail below in conjunction with the accompanying drawings.

[0054] Such as figure 1 As shown, a sound source localization algorithm based on convolutional neural network CNN, including the following:

[0055] Step 1: Use the Roomsim toolkit to simulate the indoor environment with reverberation and noise, according to the signals received by the two microphones x 1 (t) and x 2 (t), calculate the cross-correlation function of the signal received by the microphone Then perform frame-by-frame interception and feature extraction to obtain a training set to prepare for the next step of model training.

[0056] The specific steps of feature extraction are as follows:

[0057] Step 1.1: The sound source is located at the training position r s ,s=1,2,...,k, the microphone array records the signal s(t) sent by the sound source at this position;

[0058] Step 1.2: Calculate the reverberation signal x received by the microphone ac...

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 present invention provides a CNN-based sound source positioning algorithm. A method comprises the steps of introducing the characteristics for identifying a cross-correlation function, utilizing an Roomsim simulation environment and the signals received by two microphones to obtain the cross-correlation function with the reverberation and under a noisy environment, and intercepting frame by frame to obtain a training set and a test set, training the characteristics to obtain a CNN, namely a CNN model, during the training process, adopting an ReLU function as an activation function, carrying out the sound source positioning estimation on the test set under the model, and finally adopting a Bayesian decision to construct a determination formula to decide the types of the test samples, sothat a position where the conditional probability p(rs|Y) is maximum is the real position where a sound source is estimated. By the realization of the algorithm, the noise and reverberation problemsin the conventional sound source positioning are solved effectively.

Description

technical field [0001] The invention relates to a research on a sound source localization algorithm based on a neural network (CNN), and belongs to the field of information technology. Background technique [0002] For the sound source localization algorithm, how to improve the anti-noise and anti-reverberation capabilities has been the research focus for a long time. In the actual environment, when the signal-to-noise ratio is small and the reverberation is serious, the improvement measures based on traditional algorithms Hard to have a noticeable effect. In addition, when the microphone cannot receive the direct sound of the sound source, it is also difficult to locate. The present invention proposes to use a convolutional neural network (CNN) to identify a phase-transformed weighted generalized cross-correlation function (GCC-PHAT) for sound source localization. Experiments show that the convolutional neural network (CNN) has better localization performance in low signa...

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): G01S5/18G06N3/04
CPCG01S5/18G06N3/045
Inventor 万新旺王吉廖鹏程陈中倩
Owner NANJING UNIV OF POSTS & TELECOMM
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