Identification of arm movements using correlation of electrocorticographic spectral components and kinematic recordings

Abstract

The purpose of this study was to explore the possibility of using electrocorticographic (ECoG) recordings from subdural electrodes placed over the motor cortex to identify the upper limb motion performed by a human subject. More specifically, we were trying to identify features in the ECoG signals that could help us determine the type of movement performed by an individual. Two subjects who had subdural electrodes implanted over the motor cortex were asked to perform various motor tasks with the upper limb contralateral to the site of electrode implantation. ECoG signals and upper limb kinematics were recorded while the participants were performing the movements. ECoG frequency components were identified that correlated well with the performed movements measured along 6D coordinates (X, Y, Z, roll, yaw and pitch). These frequencies were grouped using histograms. The resulting histograms had consistent and unique shapes that were representative of individual upper limb movements performed by the participants. Thus, it was possible to identify which movement was performed by the participant without prior knowledge of the arm and hand kinematics. To confirm these findings, a nearest neighbour classifier was applied to identify the specific movement that each participant had performed. The achieved classification accuracy was 89%.