| With the aggravation of population aging,China is faced with the double superimposed pressure of the rapid growth of the elderly and the disabled with motor disabilities.In the face of the rapidly increasing demand for rehabilitation and nursing,it is imperative to guarantee and improve the quality of life of the elderly and the disabled through scientific and technological innovation.With the continuous innovation of new materials and artificial intelligence technology,wearable exoskeleton robot has become a feasible program for the elderly and the disabled to improve their limb motor function and life quality.Due to the weak physical mobility of the elderly with disabilities,the coordination of motion intentions between exoskeleton robots for medical rehabilitation and users is particularly important.Generally,biomechanics-based and EMG-based sensing technologies are used to identify human motion intentions.However,for those special people with mutilated limbs and muscular nervous system injury,there is no movement sensation in relevant limbs and muscles,so biomechanics-based perception technology cannot obtain the user's movement information.EMG-based sensory recognition technology can only detect very weak electrical signals,and patients with limb disabilities even can not detect signals.The introduction of brain-computer interface(BCI)technology has greatly improved the ability of motor intention acquisition of special population.BCI technology builds a communication channel between the brain and external devices,and users can directly control the exoskeleton robot to carry out rehabilitation training and assist in sports through the intention information of the brain.However,in the existing studies,the motor intention decoding ability based on motor imagery EEG signals needs to be improved,and the lower limb joint motion decoding method based on EEG signals cannot meet the needs of special populations.Therefore,aiming at the decoding of human joint motion intention based on BCI technology,this study carried out research on motion intention recognition method based on MI-EEG and reconstruction method of knee joint EMG based on EEG.Through the intention decoding of EEG signals and the mapping between EEG and EMG,reliable intention information is provided for the continuous motion control of the knee exoskeleton robot platform.Firstly,a weighted ensemble learning method based on time-frequency decomposition,and an ensemble learning method based on temporal,spatial features with multiscale filter banks are proposed.Based on the idea of ensemble learning multi-classifier decision making,an ensemble classifier model with sample distribution diversity,timefrequency diversity,and domain diversity is constructed to obtain more stable intention decoding results,thus improving the classification performance of EEG signals.Secondly,a power spectrum pattern difference-based time-frequency sub-band selection method is proposed.The frequency band and time window of ERD/ERS pattern activated by motor imagery were found by using the power spectrum pattern difference information between inter-and intra-sample sets,and then used for feature extraction and classification,which further improved the classification performance of EEG signals.Finally,an EEG-based EMG signal reconstruction method is proposed,and an EEG-EMG experimental paradigm based on the knee exoskeleton platform is designed.By constructing the mapping relationship between the EEG signal and the lower limb EMG signal,the reconstructed EMG signal can be used to decode the motion of the lower limb muscles,and provide continuous control input for the lower limb exoskeleton robot for special populations.The intention decoding method proposed in this study can improve the decoding performance of BCI,and the lower limb EMG reconstruction technique based on EEG signals builds a stable EEG-EMG mapping relationship.It provides a new research idea for BCI-based exoskeleton robot control technology. |
PDF Full Text Request