BCI-driven personalized stress detection and management solution using deep Q-network

Abstract

Brain Computer Interface (BCI) is one of the hottest topics in today’s world. Brain computer interface has been heavily used for medical domain. In BCI, Researchers are working with the Brain signals. So, they are deal with EEG signals that are non- invasive. Specially most Researchers are committed to find out various treatments to various disorders using BCI technology. But EEG signals are non-stationary. That means statistical features of the signal is changing time to time. Distribution is varying. Specially with various Artifacts such as eye blinks or some other environmental changes may be effect to the EEG signals that we are capturing. Therefore, in order to get the most accurate results of the BCI applications handling the non-stationary of the EEG signal is crucial. Stress management is also a major concern today. Having too much of stress can be effect to the mental health as well as physical health. People will gain negative thoughts when they are in stressful situation. They try to commit suicide if they unable to handle their stress level. In phycology councilors are practicing various therapy sessions for managing the stress. Music therapy is the therapy they have most commonly used and shows as a most effective way to managing the stress. And also, it has found that BCI technology is used for monitor and managing the stress level. But there is a lack of Research explored Reinforcement Learning technology to accurately detect stress from the non-stationary EEG signal. So, to fill that gap this research focuses on Reinforcement Learning (RL) technology for accurately classify stress from the non-stationary dynamic EEG signals. Reinforcement Learning in the sense specially this research uses Deep Q-Learning (DQN) which is coming under Reinforcement Learning. And finally combined machine learning for recommending the suitable relaxing music from Spotify if the stress has been detected for management of the stress. Desktop application has also been implemented in the part of this work that can be used by anyone for personalized stress management as well.