A Novel Approach to EEG Neurofeedback via Reinforcement Learning

Since the invention of EEG brain scanning technology, cognitive response modeling and brain state optimization has been a topic of great interest and value. In particular, applications for improving musical therapy via neurofeedback have shown promise for brain state optimization. Here, we propose a novel Humanistically Intelligent (HI) system where brain waves are interpreted by a real-time deep reinforcement learning agent that controls an audio modulation system in order for the user to achieve a target brain state. The modulated audio is then visualized using a simulated Sequential Wave Imprinting Machine (SWIM). In our tests comparing the proposed system to a conventional neurofeedback system, we found that the proposed system consistently led to a more meditative brain state with p = 0.06. We show that the proposed system is promising for brain state optimization tasks, advancing the intelligent utilization of EEG brain scan data in Humanistically Intelligent feedback loops.