The EEG-based Lower Limb Exoskeleton System Optimization Strategy Based on Channel Selection

Compared with traditional exoskeleton robots, the brain-computer interface (BCI) based lower limb exoskeleton system can directly control the robot by recording the user's brain activity signals, such as the electroencephalogram (EEG). The brain activity signals contain the original and global motion intention, which is capable of operating the walking assistance exoskeleton device in a way that matches the operator's will more closely. In our previous work, an EEG-based lower limb exoskeleton system and a human-machine interaction paradigm that orient to the scenario of dynamic motion is designed. However, the EEG sensor configuration is a time-consuming process and limit the utility of the EEG based system. Besides, channel redundancy makes it difficult to explore experimental mechanisms of the designed paradigm. Therefore, in this paper, a heuristic channel selection optimization algorithm with the L1 norm is applied to the BCI system. Based on the data of four subjects, we verify the performance of the selected optimal channel subset and explore a universal optimal channel subset. Tested on another two new subjects, the universal optimal subset shows a promising prospect in improving the performance of the EEG-based exoskeleton system.