Analysis of the Principle and Applications for Emotion Adding in Music Composition

As a matter of fact, emotion in music has always been essential for the audience to resonate with music. Similarly, emotion in computer generated music also follows such principle. With this in mind, this study explores different models and methods used to generate emotion in music including EEG-Based Music Emotion Prediction, a two-layer feature extraction model combined with Convolutional Recurrent Neural Network (CRNN), as well as the semi-supervised emotion-driven music generation model CDGMVAE. At the same time, the paper examines reinforcement learning-guided pre-trained models such as GPT-2 and ERoPE-Transformer. To be specific, this research evaluates the generated emotion through valence-arousal model and classification performance metrics, demonstrating their effectiveness in generating emotionally expressive music. According to the analysis, the research provided the limitation of the emotion in models and the prospects for future improvement also been discussed. Overall, these results push the boundary of analysis on emotion on emotion in computer music.