Human Motion prediction using EEG data for Assistive BMI Applications

Brain injury often leads to motor impairment, significantly affecting the patient's well-being. Several works address motor impairment, including BMIs. However, the application of BMIs is often limited by the need for a strictly controlled environment and careful calibration, making them unsuitable for daily life environments. To overcome this challenge, we explored the use of a research-grade wireless passive-electrode brain signal recorder. This device requires less calibration time and offers greater ease of use. To compensate for the reduced signal quality, we deployed compact yet powerful convolutional neural networks to decode the brain signals. Another challenge is the lack of sufficient work addressing intricate human motion, which restricts the full potential of BMIs. In this work, we aim to improve the understanding of complex human motion by focusing on finger and arm movements. The experimental results are encouraging, demonstrating a promising method to develop a system that interprets human motion through brain signals, enabling a more natural way to control a robot for motor rehabilitation.