CVAE-GAN Emotional AI Music System for Car Driving Safety

Abstract

Musical emotion is important for the listener’s cognition. A smooth emotional expression generated through listening to music makes driving a car safer. Music has become more diverse and prolific with rapid technological developments. However, the cost of music production remains very high. At present, because the cost of music creation and the playing copyright are still very expensive, the music that needs to be listened to while driving can be executed by the way of automated composition of AI to achieve the purpose of driving safety and convenience. To address this problem, automated AI music composition has gradually gained attention in recent years. This study aims to establish an automated composition system that integrates music, emotion, and machine learning. The proposed system takes a music database with emotional tags as input, and deep learning trains the conditional variational autoencode generative adversarial network model as a framework to produce musical segments corresponding to the specified emotions. The system takes the music database with emotional tags as input, and deep learning trains the CVAE-GAN model as the framework to produce the music segments corresponding to the specified emotions. Participants listen to the results of the system and judge whether the music corresponds to their original emotion.