| Summary: | 博士 === 南臺科技大學 === 電機工程系 === 105 === Some people with severe disability, for example, people with amyotrophic lateral sclerosis/motor neuron disease (ALS/MND), have difficulties with moving some parts of their bodies including walking, eating, grasping, etc. These people with severe disability need some high technology assistive devices to help them perform some daily activities such as communicating, entertainment and self-feeding. Brain-computer interface (BCI) could be one of the efficient, helpful technologies and can be utilized as a novel assistive device to help these people with severe disability to perform these daily activities.
One of the BCI related methods called steady state visual evoked potential (SSVEP) is implemented and discussed in some BCI applications including English spelling system, recreational device and robot control in this thesis. In SSVEP-based BCI applications, the subjects were asked to focus their attention on one of the several flickering visual stimuli to choose the desired command. While the subject gaze at the visual stimulus, the electroencephalography (EEG) signal of the subject is acquired and analyzed through the occipital region of the subject’s brain. The brain signal of the subject will contain the same frequency response as the selected stimulus frequency and its harmonic response. After the raw EEG signal of the subject is preprocessed and recognized by the algorithm we developed, the feature of EEG corresponding to the selected visual stimulus can be extracted and used as the input of the BCI decision model to finish recognistion and control application.
In this study, the SSVEP method was used in several BCI applications. In the SSVEP-based BCI application for spelling system, the implementation of entropy encoding algorithm with dummy character (EEA_D) can yield information transfer rate (ITR) equaling to 42.62 bits/min which is greater than the ITR of the balance structure that equals to 31.9 bits/min. In the SSVEP-based BCI application for controlling air swimmer toys, the implementation of fuzzy tracking and control algorithm as decision model yields higher recognition rate equaling to 96.97% compared to the canonical correlation analysis (CCA) as a baseline which yields 94.49%. The fuzzy feature threshold algorithm applied in SSVEP-based BCI application for controlling mobile robot can yield ITR equaling to 20.57 bits/minutes. The fuzzy decision model with fast Fourier transform (FFT) and magnitude squared coherence (MSC) input features can yield F-score equaling to 0.5889 and 0.6038 in single visual stimulus test (SVST) and multiple visual stimuli test (MVST) respectively, while the CCA baseline yielded 0.4052 and 0.4005 in SVST and MVST respectively. Based on this result, the SSVEP-based BCI using the proposed approach is successfully developed on spelling system, air swimmer toys, mobile robot, and robot with the function of meal assistance.
|
|---|
