Auxiliary Force EEG (AFEEG) Recognition Oriented to Stroke Patients

摘要

Purpose In this paper, electroencephalogram (EEG) is used to recognize the pattern of the different auxiliary forces (AFs) on the upper extremity. It is expected that the robot obtains human feedback information in order to achieve robot control based on EEG. Method Collect the EEG of ten healthy subjects under different AF states, which is divided into non-AF, AF1, and AF2. Combined with power spectrum method, WPT is used to determine the key frequency bands and the EEG energy spectrum distribution state of key signal channels in different frequency ranges. Then, extract the energy features using support vector machine to classify the EEG data. Finally, ant colony algorithm is applied to optimize the key channels. Result According to the energy spectrum distribution of EEG at different frequency bands, it can be concluded that the frequency bands of the correlation features were 7.33-7.81 Hz and 8.30-8.79 Hz. The correct recognition rates were 84.54 and 79.43% for the two-stage classification and multi-stage classification, respectively. Through the optimization of the relevant electrode channels for assisted brain electricity, the original 13 channels were compressed to within 10 channels. Conclusion The use of EEG enables offline recognition of the magnitude of the AF on the upper extremity. During the rehabilitation training, according to patient's active intention based on EEG signals, robot exerts different auxiliary forces. It is expected to provide an important technical approach for the rehabilitation robots to exert different auxiliary forces on patients in different rehabilitation periods.