Supervised dictionary learning audio classification method and system based on data driving, and medium

A dictionary learning, data-driven technology, applied in the field of sparse representation, to achieve the effect of improving pairwise orthogonality, excellent generalization ability, and excellent performance

Pending Publication Date: 2021-12-17
SOUTH CHINA UNIV OF TECH
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Benefits of technology

This patented technology allows for efficient training dictionaries that are more accurate than existing methods like histogram or spectral analysis techniques. It also has improved accuracy when applied across different scenarios where sound signals may vary greatly due to factors like background noise levels. Overall, this new technique helps create better quality speech recognizers with high confidence scores at various stages during security testing processes.

Problems solved by technology

This patented technical problem addressed in this patents relates to improving dictionaries trained with data from an unstructured dataset such as speech or image signals without losing important parts like sound quality due to noise interference during transmission over communication networks.

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
  • Supervised dictionary learning audio classification method and system based on data driving, and medium
  • Supervised dictionary learning audio classification method and system based on data driving, and medium
  • Supervised dictionary learning audio classification method and system based on data driving, and medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0050] Such as figure 1 As shown, the present embodiment provides a data-driven supervised dictionary learning audio classification method, comprising the following steps:

[0051] S1. Determine the number of categories C of the sample set, and use the input sample x n , and its corresponding class label y n Train C class-specific dictionaries D c , c∈[1,C], such as figure 2 As shown, it specifically includes the following steps:

[0052] S11. Initialize dictionary D c 0 , learning rate η 0 , the learning rate update rate α, the number of iterations T;

[0053] S12. Determine the loss function J;

[0054] Furthermore, the specific form of the loss function J is:

[0055] J(A,D)=J 1 (D, A)+μJ 2 (D, A)+λJ 3 (A)+γ 1 J 4 (A)+γ 2 J 5 (D);

[0056]

[0057]

[0058]

[0059]

[0060]

[0061] Among them, μ is the sample constraint parameter, λ is the classifier constraint parameter, γ 1 Constraint parameter for sparse coding, γ 2 Learn constraint p...

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 supervised dictionary learning audio classification method and system based on data driving and a medium. The method comprises the steps of determining a sample set category number; training a specific class dictionary by using the input samples and the corresponding class labels; obtaining a sparse code of the input sample by using the trained dictionary, and training an SVM classifier by taking the sparse code as a feature; and classifying the input samples by using the trained dictionary and the trained SVM classifier, and outputting a prediction label. The intra-class uniformity is minimized and the class separability is maximized by learning one dictionary through each class, the sparsity is improved so as to control the complexity of decomposition of signals on the dictionary, meanwhile, class-based reconstruction errors are minimized, and the pairwise orthogonality of the dictionary is improved. The invention can be widely applied to a plurality of scenes, such as computational auditory scene recognition and music chord recognition; and the test on the data set is relatively stable, and the generalization ability is excellent.

Description

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

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
Owner SOUTH CHINA 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