Key points are not available for this paper at this time.
With the rapid development of artificial intelligence in the fields of music generation and emotion computing, how to achieve accurate matching between music content and user emotions has become a research hotspot. However, existing methods still have shortcomings on emotion recognition accuracy, emotional consistency in melody generation, and multidimensional evaluation feedback. Therefore, a emotion-driven music generation and multidimensional evaluation model based on artificial intelligence is developed. The model extracts Electroencephalogram (EEG) signals, integrates the main frequency feature weighting mechanism and lightweight neural network structure to achieves efficient emotion recognition of multi-frequency EEG signals. Genetic algorithm is used for initial melody construction and combined with long short-term memory networks to optimize the temporal sequence of melodies. Meanwhile, a multidimensional evaluation feedback structure is introduced to strengthen the consistency regulation between melody and emotional labels. Based on two classic datasets, the research sets up four types of emotion generation simulation scenarios and compares them with three mainstream music generation models. The experimental results showed that the model exhibited advantages in key indicators such as note prediction accuracy, rhythm matching, and style similarity. The melody similarity score reached 0.871, with a subjective evaluation mean of 4.6 points. The research results indicate that the model not only has high generation accuracy and emotional expression, but also exhibits good robustness and generalization ability in multiple scenarios, providing an efficient intelligent solution for emotional music generation and human–computer interactive creation.
Liping Li (Wed,) studied this question.