Environmental sound recognition method and device based on hybrid multi-task learning
A multi-task learning and environmental sound technology, which is applied in neural learning methods, character and pattern recognition, speech analysis, etc., can solve the problems that hinder the application of multi-task learning methods and the high cost of data sample preparation, and achieve low cost and fast generation speed , the effect of improving performance
Active Publication Date: 2021-04-20
SOUTH CHINA NORMAL UNIVERSITY
View PDF6 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Both of these methods require a lot of manpower, making the preparation of data samples very expensive, which hinders the application of multi-task learning methods in the field of environmental sound recognition.
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0066] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention.
[0067] In some processes described in the specification and claims of the present invention and the above-mentioned drawings, a plurality of operations appearing in a specific order are contained, but it should be clearly understood that these operations may not be performed in the order in which they appear herein Execution or parallel execution, the serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. Additionally, these processes can include more or fewer operations, and these operations can be performed sequentially or in parallel. It should be 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
Login to View More Abstract
The invention discloses an environmental sound recognition method and device based on hybrid multi-task learning. The method includes: acquiring a sound scene audio data set and a sound event audio data set; generating a corresponding first audio data set according to the audio data in the sound scene audio data set A sound spectrum atlas, generating a corresponding second sound spectrum atlas according to the audio data in the sound event audio data set; combining the first sound spectrum atlas and the second sound spectrum atlas to obtain a mixed sound spectrum atlas; using the mixed sound Spectrum Atlas trains the constructed multi-task learning network model to obtain pre-trained model parameters; adjusts the network structure of the multi-task learning network model to obtain a single-task learning network model; The learning network model is initialized, and the single-task learning network model is tuned and trained using the first sound spectrum atlas to obtain the final model for environmental sound recognition.
Description
technical field [0001] The present invention relates to the technical field of audio recognition, and more specifically, to an environmental sound recognition method and device based on hybrid multi-task learning. Background technique [0002] Environmental sound recognition technology has great application potential in security monitoring, smart home, multimedia retrieval and other fields. It perceives environmental semantic information by analyzing audio data recorded in real-life environments, mainly including different tasks such as sound scene classification, sound event recognition, and audio labeling. These tasks all use real-life audio data, but are labeled differently depending on the intelligent computing task. Therefore, these tasks are highly correlated in learning, for example, multiple tasks can share a certain audio feature pattern (for example, the sound of a vacuum cleaner in a domestic environment has similar characteristics to the sound of a machine runni...
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
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G10L25/51G10L25/30G06N3/04G06N3/08G06K9/62
CPCG10L25/51G10L25/30G06N3/08G06N3/047G06N3/045G06F18/2415
Inventor 郑伟平蒋大灿
Owner SOUTH CHINA NORMAL UNIVERSITY