Machine learning for electroencephalography decoding and robotics dextrous hands movement

Abstract

This work focuses on using machine learning (data analysis) for interpretation and understanding of brainwaves resulting from electroencephalography during a grasping task. Electroencephalography — EEG — was used for acquisition of brain neural signals thought activity, hence to layout a control strategy for robotic hand and fingers movements. This is done via decoding, in real-time, the neural activity associated with fingers motions. Results are used for training robotics dexterous hands, and might allow people with spinal cord injury, brainstem stroke, and ALS (amyotrophic lateral sclerosis) to control a robotic-prosthetic by thinking about movements. The project is novel in a sense, it relies on detecting grasping features for a human grasping using Principle Component Analysis (PAC), hence to learn these features for recognitions applications.