Classification of classical music genres based on Mel-spectrogram and multi-channel learning

Music genre classification has become a major focus of study as audio processing. Mel-spectrogram and multi-channel learning, MC-MelNet, is proposed in this work for the categorisation job of classical music genres. Combining the Mel-spectrogram and other audio characteristics, with a multi-channel learning framework, the model performs thorough modelling of audio signals. Complete use of the multidimensional information in the audio data enhances the categorisation accuracy. By means of end-to-end training, MC-MelNet simplifies the conventional feature engineering processes and simultaneously performs well in the tests, so attaining higher accuracy, precision, recall, and F1 socre than in the conventional approaches, which show the robustness and efficiency of multi-channel learning in the classification of classical music. The experimental results reveal that the MC-MelNet model can give significant support for the domains of audio classification and music information retrieval in the categorisation of classical music genres.