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

An Acoustic Scene Clustering Method by Jointly Optimizing Deep Transformation Features and Clustering Process

A technology of joint optimization and clustering method, applied in speech analysis, instrumentation, etc., can solve problems such as inability to obtain clustering results of acoustic scenes

Active Publication Date: 2021-03-30
SOUTH CHINA UNIV OF TECH
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The current unsupervised sound scene classification method generally separates audio feature extraction and sound scene clustering, and cannot obtain optimal sound scene clustering results.

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
  • An Acoustic Scene Clustering Method by Jointly Optimizing Deep Transformation Features and Clustering Process
  • An Acoustic Scene Clustering Method by Jointly Optimizing Deep Transformation Features and Clustering Process
  • An Acoustic Scene Clustering Method by Jointly Optimizing Deep Transformation Features and Clustering Process

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0065] like figure 1 as shown, figure 1 It is a flowchart of an embodiment of an acoustic scene clustering method for joint optimization of deep transformation features and clustering process, which mainly includes the following processes:

[0066] S1. Extract logarithmic mel spectrum features: pre-emphasize, frame, and window the audio samples of various sound scenes, and then extract the logarithmic mel spectrum of each audio frame;

[0067] S2. Initialize various types and convolutional neural networks: use each sample as an initial class, initialize and generate a convolutional neural network for extracting deep transformation features;

[0068] S3. Update the convolutional neural network and extract new deep transformation features: update the convolutional neural network parameters according to class labels and various samples, and use the updated convolutional neural network to extract deep transformation features of various samples;

[0069] S4. Merge the two most si...

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 an acoustic scene clustering method for joint optimization of deep transformation features and the clustering process. The acoustic scene clustering method comprises the following steps that a, logarithmic Mehr spectrum features of all samples are extracted and serve as an initial class, and a convolutional neural network is initialized; b, the logarithmic Mehr spectrum features of the samples are input into the convolutional neural network, and the deep transformation features are extracted; c, two most similar classes are combined through a agglomerative and hierarchical clustering algorithm, new class tags and samples are obtained and used for updating the convolutional neural network, one is subtracted from the class number, and the logarithmic Mehr spectrum features of the samples are transformed into the deep transformation features through the updated convolutional neural network; d, if the current class number is equal to the true class number, clustering is stopped, the jointly optimized acoustic scene clustering result and the convolutional neural network are obtained, otherwise, the step c is carried out. According to the method, deep transformation feature extracting and clustering are alternately carried out, the joint optimization result is obtained, and compared with a traditional clustering method, the performance is better. Compared witha traditional classification method, the method has higher universality.

Description

technical field [0001] The invention relates to the technical field of audio signal processing and pattern recognition, in particular to an acoustic scene clustering method for joint optimization of deep transformation features and clustering process. Background technique [0002] Acoustic scene clustering (ASC) is to compare the similarity of collected audio samples of various acoustic scenes, and merge the audio samples of the same category together. The purpose of acoustic scene clustering is mainly to make the machine more intelligent, so that it has the ability to distinguish the surrounding acoustic environment similar to humans, so as to provide more intelligent services for humans. Acoustic scene clustering technology is an important basis for audio monitoring, automatic assisted driving, multimedia content analysis and retrieval and other application fields, and has important research value and practical significance. [0003] The traditional supervised acoustic sc...

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 Patents(China)
IPC IPC(8): G10L25/24G10L25/30G10L25/45G10L25/51
Inventor 李艳雄刘名乐王武城张聿晗
Owner SOUTH CHINA UNIV OF TECH
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