A Novel Modular and Wearable Supernumerary Robotic Finger via EEG-EMG Control with 4-week Training Assessment

Abstract

The Supernumerary Robotic Limbs (SRL) is an emerging kind of wearable robot to help reconstruct and enhance human movement functionality by adding extra limbs, such as arms, legs, or fingers. For the control, EEG can transmit human intentions to control external devices independently, hence it has the particular advantages to control the SRL naturally especially when human limbs are occupied. However, the classification accuracy of EEG is very limited. In this paper, we proposed a novel EEG-EMG strategy to improve the classification accuracy. In addition, a novel supernumerary robotic finger (SRF) system is built with a modular design consideration. The whole system contained seven modules: EEG acquisition, EEG control, EMG acquisition, SRF control, SRF finger, TENS feedback and status information Module. To the best of our knowledge, this is the first wearable brain control SRL system. Meanwhile, as the SRF is a soft finger, the system is compliance, affordable, wearability, modularity, and lightweight, named CAWML-SRF. The finger weight is less than 195 g, and the whole system weight is less than 1 kg. The experiments show that the CAWML-SRF can enhance the grasp functionality for helping the single hand to accomplish the bimanual grasp task (such as opening the bottle and grasping the larger objects), and assist the remaining functionality to accomplish the grasp task. A 4-week EEG triggering training experiment is conducted to further evaluate the rehabilitation application potentials: the high control accuracy rate (94.53% ± 0.044) and fine learning performance of human brain to the system are verified.