Decoding hand trajectory of ECoG signals using deep learning method

This research addresses the significant challenge of limited mobility in individuals affected by limb amputation or spinal cord injuries. The study focuses on Brain-Machine Interface (BMI) systems, which decode neural activities to interpret users' intentions, fostering increased independence in performing tasks. Furthermore, BMI systems hold potential for technological advancements and improving the quality of life for broader populations.The primary objective is to decode continuous three-dimensional hand position by employing electrocorticography (ECoG) signals recorded from the motor cortex. Notably, the utilization of ECoG signals for estimating hand position in primates is crucial, given their long-term recording capabilities and access to comprehensive datasets. The research incorporates mathematical modeling, feature extraction, and estimation of motor activities based on the analyzed ECoG signals. Challenges include enhancing accuracy and computational efficiency compared to prior investigations. The proposed approach utilizes partial least squares (PLS) regression as the decoding method, demonstrating highly accurate estimation of movements with a notable average correlation coefficient of ۰.۷۱۸, effectively predicting motion trajectories compared to actual measurements.

Keywords:

Brain-machine interfaces , ECoG signals , mathematical modeling , partial least squares regression , feature extraction , decoding