著者・編者Wilaiprasi, Theerawit
Wilaiprasitporn, Theerawit
並列タイトル等A Study on Visual Stimulus Increasing Event-Related Potentials and Computer Interfaces Using EEG
一般注記Brain--computer interfaces (BCIs) have been actively researched for over two decades. One of the primary goals is to create a non-muscular communication channel for locked-in patients. Electroencephalography (EEG) is a non-invasive technique that is commonly used in BCI measurement systems. Even though BCIs have a long history, their performance is still limited by the low signal-to noise ratio of EEG. A state-of-the-art BCI application is P300-based BCI. P300 refers to a major event-related potential (ERP) component that peaks around 300 ms after visual stimulus. P300 is an electroencephalographic correlate of target recognition in decision-making tasks. The P300 is used in several brain-computer interfaces (BCIs) as a non-motor signal of decisions, such as letter choice in the P300-Speller utility. Accuracy in choice specification depends on the difference in P300 amplitude evoked by target versus non-target stimuli. In this study, I describe visual stimulus factors, color, motion-modulated, complexity-modulated and orientation-modulated, all of which enhance the difference in P300 magnitude between target and non-target stimuli for P300-based BCIs. Stimulus arrays incorporating these visual factors may be used for the design of improved P300-based BCIs with greater choice accuracy and speed.
To demonstrate advantage of research findings from visual factor studies, I report the development of a personal identification number (PIN) application using a P300-based BCI. I focus on visual stimulation design for increasing the evoked potential in the brain. Single-channel electroencephalography and a computationally inexpensive algorithm are used for P300 detection. Experimental results showed that my proposed stimulus induces higher P300 amplitude than does a conventional stimulus. For a performance evaluation, I compare two versions of the proposed application, which are based on my `original P300 BCI' and `adaptive P300 BCI'. In the adaptive P300 BCI, I introduce a novel algorithm for P300 detection to improve the information transfer rate while maintaining acceptable accuracy. Experiments with 10 healthy participants reveal that the original P300 BCI achieves mean accuracy of 83.5 % at 11.4 bits/min and the adaptive version achieves mean accuracy of 86.0 % at 18.6 bits/min.
On the basis of BCI and PIN application, I expand my research to hybrid BCI. Here, I propose a hybrid brain/blink computer interface based on a single-channel EEG amplifier. Eyelid closing and hard blink are selected as two possible inputs for control of the interface. A 2-min calibration is required before starting to use the interface. An algorithm for feature extraction and classification is developed for EEG signals from eyelid closing, hard blink, and resting. To evaluate the performance of the interface, I incorporate it into a personal identification number (PIN) application, in both visual and auditory modes. Experiments with 5 healthy participants reveal that the PIN application based on the interface achieved a mean accuracy of 97.4 %. In conclusion, I expect that my investigation will contribute to hybrid brain-computer interface research and technologies in the near future.
identifier:oai:t2r2.star.titech.ac.jp:50359226
一次資料へのリンクURLhttp://t2r2.star.titech.ac.jp/rrws/file/CTT100736171/ATD100000413/thesis_15D54102.pdf （fulltext）
連携機関・データベース国立情報学研究所 : 学術機関リポジトリデータベース（IRDB）（機関リポジトリ）