FML-based Machine Learning Tool for Human Emotional Agent with BCI on Music Application

This paper presents a human emotional agent based on a fuzzy markup language (FML) tool and machine learning mechanism for music application. The human emotional agent contains a brain-compute-interface (BCI) preprocessing mechanism, a knowledge base, a rule base, a machine learning mechanism, and robots, for music application. We first select some famous songs with various languages, including English, Japanese, Chinese, and Taiwanese, and invite some learners to wear a wireless BCI device to listen to the chosen songs under the designed experimental modes and models. After listening, the learners help label their real-time feelings of each song. We analyze the learners' physiological indices transformed from the collected brainwaves to observe whether the learners could have mind communication with the robot. The experimental results show that the FML-based machine learning tool can help the proposed human emotional agent effectively on the music application. They also show that there is a relationship between the physiological indices and the learners' culture, familiarity, and preference for the listened songs.