Music Emotion Analysis Based on Deep Learning Techniques for Streaming Platforms

Abstract

The purpose of music therapy is to help listeners explore self-emotions and experiences, reduce pain, soothe mood, and increase motor coordination through their responses to music. Many arguments deriving from the perspectives of thalamic function, endocrine volume, β-endorphin, and psychology strengthen the scientific status of music therapy. Using the Chinese music library from 1922 to 2021 provided by Spotify and the different music audio characteristics, we analyzed the melody style of songs in the library to train an artificial intelligence model and features of music emotion. In addition, we also analyzed and compared the effectiveness and results between machine learning and deep neural network after applying them to music emotion classification. According to the experimental results, the tagging of Chinese music assists the listener to understand the song. The feeling and memory triggered by the patient’s connection to the songs or the selection of music in appropriate occasions steer patient’s physical, mental, and emotional states toward the desired direction for therapeutic purposes.