| Summary: | 碩士 === 中原大學 === 機械工程研究所 === 103 === BCIs (Brain-Computer Interface) used to help people who suffered for severe motor–disabilities (such as stroke or ALS patients) to interact with external world or control devices. In recent years, many BCIs for stroke rehabilitation with motor imagery have been report.
To develop a BCI system for rehabilitation, this thesis designs an experimental procedure based on lower limb motor imagery. The experiment consists of three states, including resting state, left foot motor imagery and right foot motor imagery. Participants are divided into three groups, stroke, healthy young and healthy elderly, respectively. This thesis analyzes the classification rate of unilateral foot motor imagery and resting state, brain topography, and use results of statistics to find EEG difference between young and elderly, sound and affected side, stroke patients and healthy elderly.
Feature extraction methods include Band Power (BP), Common Spatial Pattern (CSP). This thesis employs K-nearest neighbor (K-NN) and Support Vector Machine (SVM) as classifier. The results show that CSP feature extraction method combined with SVM classifier can obtain the best classification rate between unilateral foot motor imagery and resting state. In the group of healthy young, classification rates are 89.94% (left vs. resting) and 88.94% (right vs. resting). In the group of healthy elderly, classification rates are 91.88% (left vs. resting) and 88.06% (right vs. resting). In the group of stroke, classification rates are 87.19% (affected side vs. resting) and 85.56% (sound side vs. resting).
According to brain topography and the results of statistics, many significant differences between healthy young and healthy elderly in left foot motor imagery and right foot motor imagery are found. It indicates that lower limb motor imagery has age effect. Compare the stroke sound side with healthy elderly, 8-12 Hz have more significant differences, many electrode sites show significant differences in the frequency band of 8-12 Hz. There are no significant differences in period of 0~2 s. Compare the stroke affected side with healthy elderly, 16-20 Hz have more significant differences, and lots of bands show significant differences in electrodes FZ、CZ、CPZ. There are no significant differences in period of 0~2 s. Summarizing all the results, development of a low limb motor imagery-based BCI rehabilitation system for stroke patients is possible in the future.
|
|---|
