Channel self-coordinated attention and time-frequency masking data augmentation based music genre classification
计算机科学辑要 · 2026-10-20 · 1 (3) · 43
DOI: 10.63313/cs.8027 ↗At a glance
Journal article in 计算机科学辑要, published online on 2026-10-20. Written by Zhifang Wen. 22 references. The full text is freely available (open access, CC BY-SA 4.0).
Abstract (original language)
Music genre classification (MGC) is an indispensable part of music information retrieval. Because convolutional neural network (CNN) has made a breakthrough in computer vision, researchers apply it to MGC. When the CNN captures local features, the convolution kernel treats all channels as a whole for feature extraction, however, the feature information of each channel is not equally important. Therefore, this paper proposes channel self-coordinated attention (CSCA) to adaptively calibrate the contribution of different channels to the global information in the feature map. At the same time, in order to solve the model overfitting, this paper adopts the method of randomly masking the time-frequency domain of the spectrogram for data augmentation. Experimental results show that the accuracy of our model is 90.8% on the GTZAN dataset.
Cite
APA
Wen, Z. (2026). Channel self-coordinated attention and time-frequency masking data augmentation based music genre classification. 计算机科学辑要, 1(3), 43. https://doi.org/10.63313/cs.8027
Vancouver
Wen Z. Channel self-coordinated attention and time-frequency masking data augmentation based music genre classification. 计算机科学辑要. 2026;1(3):43. doi:10.63313/cs.8027
BibTeX
@article{Wen2026,
author = {Wen, Zhifang},
title = {Channel self-coordinated attention and time-frequency masking data augmentation based music genre classification},
journal = {计算机科学辑要},
year = {2026},
volume = {1},
number = {3},
pages = {43},
doi = {10.63313/cs.8027}
}
EUVIMED