A hybrid BCI system based on motor imagery and transient visual evoked potential

• Published in: Multimedia tools and applications 2020-04, Vol.79 (15-16), p.10327-10340
• Main Authors: Feng, Zhengquan, He, Qinghua, Zhang, Jingna, Wang, Li, Zhu, Xinjian, Qiu, Mingguo
• Format: Article
• Language: English
• Subjects: Classification; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Electroencephalography; Feature extraction; Feature recognition; Feedback; Human-computer interface; Hybrid systems; Image segmentation; Medical imaging; Motivation; Movement; Multimedia Information Systems; Muscles; Neural networks; Rehabilitation; Special Purpose and Application-Based Systems; Synchronism; Time domain analysis; Training; Visual signals; Visual stimuli; Wavelet analysis
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-019-7607-3

Motion imaging (MI) refers to the psychological realization of motions without movement or muscle activity; the basis of neural rehabilitation as a brain-computer interface (BCI) technique has been extensively studied. The combination of motor imaging and brain-computer interface technology can take advantage of patients’ willingness to take the initiative to assist them in rehabilitation. Studies have shown that MI combined with BCI rehabilitation training is better than traditional rehabilitation training. Transient visual evoked potentials and motor imaging constructed a hybrid BCI system. Three healthy subjects were tested. EEG signals were superimposed preprocessing according to visual stimulus superimposed frequency and motor guidance frequency respectively. Transient visual evoked EEG segmentation is used as a control signal of choice, the use of wavelet decomposition helps to extract features, and then use BP neural network recognition for classification and identification. Visual guidance, motion-oriented event-related synchronization, or desynchronization feature signals as rehabilitation exercise control signals, are using time-domain sliding energy analysis to extract features, and then using BP neural network recognition for classification and identification. EEG signals collected in the experiment were superimposed signals of transient visual evoked and motorized EEG. There were 300 transient electroencephalogram (EEG) and 100 segments Imagine EEG segmentation. According to the results of the test, the average recognition rate of visual evoked EEG reached 95.42%; the average recognition rate of motor imaginary EEG was 73.08%, but there was a large individual difference in motor imaging EEG signals except 1 Name of the test rate of 85%, the remaining two subjects were less than 70% recognition rate. There is a large individual difference between motion imaging and signal feature recognition, and it takes a long time to train. Therefore, it is necessary to study further the selection of control signals for rehabilitation training. As the threshold feedback signal, controlling the amplitude feedback of rehabilitation training can promote the motivation of participants’ motivation to stimulate and enhance the rehabilitation treatment effect.