Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Sound event detection method based on convolutional neural network

A technology of event detection and sound, which is applied in the field of sound event detection based on convolutional neural network, can solve the problems of limited detection accuracy, large number of deep neural network parameters, and high computational complexity, so as to reduce the number of parameters and computational complexity, The effect of reducing power consumption and computational complexity and reducing detection accuracy

Active Publication Date: 2020-11-13
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF16 Cites 25 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] How to improve the detection accuracy of SED is the focus of the existing SED technology research, but the detection accuracy of the sound event detection method based on traditional machine learning is limited; while the sound event detection method based on neural network can improve the detection accuracy, but There are limitations: that is, the number of deep neural network parameters is large, the calculation complexity is high, the storage space required is large, and the power consumption is large, which makes it unsuitable for IoT devices with severe power consumption and resource constraints.

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
  • Sound event detection method based on convolutional neural network
  • Sound event detection method based on convolutional neural network
  • Sound event detection method based on convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0055] Build the neural network model in the LCSED model of this embodiment, that is, set the network structure of the mixed convolution module, the lightweight double attention mechanism module, the time series fully connected layer and the aggregation layer:

[0056] Among them, the convolution kernel of the convolution layer B_Conv of the hybrid convolution module is set to 3×3×64, where 3×3 represents the convolution kernel band, and 64 represents the number of channels; the input dimension of the convolution layer B_Conv is (240, 64,1), the output dimension is (240,64,64); in the input and output dimensions of each network layer, if it is in the form of (A,B,C), then (A,B) represents the input / output characteristics The size of the graph, where A represents the number of audio frames, B represents the feature dimension; C represents the number of channels of the feature map;

[0057] The 4 densely connected convolutional blocks of the hybrid convolution module all include...

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 sound event detection method based on a convolutional neural network, and belongs to the technical field of audio processing. The method comprises the steps: firstly, performing primary feature extraction on an audio stream; then sending the extracted primary feature to a neural network for feature extraction and classification of sound events, and finally obtaining prediction probabilities of various sound events; and if the prediction probability of the current type of sound event exceeds a preset classification threshold, considering that a corresponding sound event exists in the current audio stream. A sound event detection model is small in parameter quantity and low in calculation complexity, so that the power consumption and the calculation complexity of the related Internet of Things equipment during sound detection processing are greatly reduced; and the detection precision equivalent to that of an existing sound event detection model is maintained. Therefore, the sound event detection method provided by the invention can be effectively applied to embedded intelligent equipment and the like.

Description

technical field [0001] The invention belongs to the technical field of audio processing, and in particular relates to a sound event detection technology based on a convolutional neural network. Background technique [0002] Sound event detection means that the device detects one or more types of sound events (collectively referred to as multiple sound events) that exist at the current moment from the continuous audio stream. Sound event detection (SED) technology has been widely used in smart home, video surveillance, environmental monitoring and other fields. For example, in a smart home application, SED technology can be used to detect a baby crying and notify the parents in the kitchen. In video surveillance applications, SED technology can be used to trigger video surveillance when abnormal sound events such as gunshots or screams are detected, thereby realizing video surveillance driven by sound events and greatly reducing power consumption. In environmental monitorin...

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): G10L25/78G10L25/30G10L25/03G10L15/16G10L15/08G06N3/04
CPCG10L25/78G10L25/30G10L25/03G10L15/08G10L15/16G06N3/045
Inventor 周军杨明雪
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products