Music Genre Classification Based on Visualized Spectrogram

Abstract

In recent year, society is faster than before, and many people cannot be properly released under high pressure, resulting in a steadily increasing suicide and crime rate. To prevent the increasing pressure index, it is necessary to properly release the pressure. Therefore, how to release the pressure quickly and conveniently is an important issue for today’s society. In order to improve the accuracy of the therapist’s selection of the correct music genre when performing music therapy, we propose the use of a visual spectrogram, and apply the YOLOv4 for "Spectrogram based on Mel spectrum using MFCC conversion", and using mAP as our scoring standard. We get the average accuracy by 10 times experiments, and obtain 92.4% in the highest mAP value, and 91.33% in the average mAP value.