Classification of hand posture from electrocorticographic signals recorded during varying force conditions

In the presented work, standard and high-density electrocorticographic (ECoG) electrodes were used to record cortical field potentials in three human subjects during a hand posture task requiring the application of specific levels of force during grasping. We show two-class classification accuracies of up to 80% are obtained when classifying between two-finger pinch and whole-hand grasp hand postures despite differences in applied force levels across trials. Furthermore, we show that a four-class classification accuracy of 50% is achieved when predicting both hand posture and force level during a two-force, two-hand-posture grasping task, with hand posture most reliably predicted during high-force trials. These results suggest that the application of force plays a significant role in ECoG signal modulation observed during motor tasks, emphasizing the potential for electrocorticography to serve as a source of control signals for dexterous neuroprosthetic devices.