Estimation of Virtual Force Sensation in Vibrotactile Stimulation during Grasping Tasks Using Machine Learning on Electroencephalogram Data

Sensory impairment arises from the compression or damage of sensory nerves, leading to difficulties in grip control as symptoms progress. To quantitatively assess peripheral nerve sensory function in patients with sensory impairment, we developed a vibration device that presents virtual haptic sensations using an external vibration actuator. This system generates virtual forces in the forward, backward, left, and right directions by stimulating the vibration sense with asymmetric waveforms while holding the device. However, the perceived direction of traction is determined subjectively by the participant, which poses a challenge for qualitative assessment. In this study, we aimed to clarify sensory responses independent of participant subjectivity by measuring and analyzing human cortical activity. We investigated the feasibility of estimating vibration stimulus of virtual forces using Electroencephalogram (EEG) through machine learning. Two types of motor tasks were performed: a high-stimulation condition and a low-stimulation condition. We found that the highest discrimination rate was observed in gamma waves under the high condition, suggesting that the direction of traction presented by the vibration device can be estimated from gamma waves and that the intensity of the haptic sensation depends on the discrimination rate.