Robustness of musical features on deep learning models for music genre classification

作者：

Highlights：

• Robustness of musical features analyzed for music classification using DL models.

• Mel-based features and Swaragram generally perform well across various music styles.

• Large frame and small sliding window length envelopes are suitable for DL models.

• Small DL models suffer from underfitting, whereas large overfit on musical features.

摘要

•Robustness of musical features analyzed for music classification using DL models.•Mel-based features and Swaragram generally perform well across various music styles.•Large frame and small sliding window length envelopes are suitable for DL models.•Small DL models suffer from underfitting, whereas large overfit on musical features.

论文关键词：Musical features,Music genre classification,Music information retrieval,Deep learning

论文评审过程：Received 27 July 2021, Revised 7 March 2022, Accepted 10 March 2022, Available online 19 March 2022, Version of Record 28 March 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116879