Research On Key Techniques Of Signal Processing And Control In Motor Neural System Reconstruction By BCICFES

Functional Electrical Stimulation (FES), one of the most frontier areas and direction in neuron engineering research and rehabilitation Medicine, reconstructs the peripheral nervous system externally and restores the muscular motor function for patients with paralyzed limb result from spinal cord injury. However, lack of convenient approach to transmit subjective intention of patients to control the FES device directly restricts the application of FES seriously that make it a technical bottleneck for further popularization as its complex, bad self adaptability and interfered easily.Brain-Computer Interface (BCI) as a rising technique developed in recent years makes it possible for patients with paralyzed limbs to control the FES device by the their intention personally. It can be a new trend in neural rehabilitation engineering research to reconstruct the peripheral nervous system by BCICFES which control FES for rehabilitation training by monitoring the movement intention using BCI technique. This paper focuses on the key techniques of signal processing and control in motor neural system reconstruction by BCICFES, including BCI denoising method, feature extraction algorithm, pattern recognition as well as the stimulation pattern and feedback control stratagem of FES system. Decoding the motor intention spatially by Extreme Energy Difference (EED) and Common Spatial Pattern (CSP), eliminating scalp EMG contamination from EEG signals by using power changes in the higher frequency bands to estimate and remove EMG contamination in the lower frequency bands, and building FES feedback control system by adaptive PID controller that modulated by Back Propagation (BP) and Radial Basis Function (RBF) Neural Network, can provide worthy technical support for the construction of BCICFES external neural system.In this study an evoked EEG experiment on different imaginary tasks of upper extremity was designed and operated, the experimental results indicated that the best rate of accuracy which gotten from CSP and RFE could be 92.86% for all the six subjects and the averaged accuracy rate was 86.22%. The proposed EMG correction method was proved a successful method for EMG noise removal while did not washed away any true EEG information. The adaptive PID controllers modulated by BP and RBF Neural Network enhanced control precision of knee joint angle significantly in tracing tasks for FES system, and achieve satisfy effect on overshoot and oscillation suppression.The results showed that the key techniques described above improved the recoginition ability for the BCICFES external neural system to predict the motor intention of the subjects and increased its capability of self adaptive and interference suppression, which were proved worthy technique supports for further development. This research laid a solid foundation for accomplishing voluntary movement rehabilitation, enhancing independent ability and achieving a satisfying treatment for the patients.