Music Response Based on Real-time Facial Expression Recognition

Abstract

This research employs information and communication technology (ICT) to develop an App that plays music lists based on emotion. Music can regenerate brain cells and ease emotions, and when users are in a negative mood, the App will automatically play the appropriate music list. Emotions are the results of users' facial expression recorded by camera and predicted by deep learning model. The accuracy of each expression is as follows: 84% for happiness; 83% for surprise; 58% for anger, 55%; sadness and 58% for neutral. The dataset is the public data set fer2013 of the competition held by Kaggle in 2013, and some data are downloaded from the Internet, such as family members who mourned after the earthquake, families of victims of terrorist attacks, and weddings. Coupled with Internet of Things (IoT) technology, this study allows users to ease their emotions through music when they are depressed, so as to avoid any improper transfer of physical or mental suffering.