Family activity sound event classification method based on additive spacing capsule network
The invention provides a family activity sound event classification method based on an additive spacing capsule network, and the method comprises the steps: studying classification of unknown types ofaudio samples based on an additive spacing capsule network model; firstly, processing to-be-trained...
Saved in:
Main Authors | , , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
07.04.2020
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention provides a family activity sound event classification method based on an additive spacing capsule network, and the method comprises the steps: studying classification of unknown types ofaudio samples based on an additive spacing capsule network model; firstly, processing to-be-trained audio output to create sample pairs, taking a logarithm Mel spectrogram signal category vector as asample; using a capsule neural network model as a backbone; constructing an additive spacing capsule network model by using a Translation layer in combination with the capsule neural network model and an additive spacing SoftMax loss function, optimizing parameters of the additive spacing capsule network model by using a gradient descent method, and finally classifying unknown category samples byusing the optimized model; the method provided by the invention has the advantages of high efficiency, strong stability, high precision and the like, can meet the basic requirements of household activity sound event classifica |
---|---|
Bibliography: | Application Number: CN201911145336 |