| Summary: | 碩士 === 國立成功大學 === 機械工程學系碩博士班 === 91 === For the totally paralyzed patients, if their brain cortex is not damaged, they might be able to control the prosthesis with the m wave. Before the m wave can be taken as a control source, we must detect the variations of m wave very precisely. However, in the primary sensory-motor area local m waves and a waves originating form the occipital cortex are detected simultaneously.
The objective of this thesis is to construct a brain-computer interface to detect the variation of the brainwave of the C3 and C4 area. Some algorithms are utilized to reduce the effect of a waves on m wave identification. Two methods, namely, Lapacian operation and adaptive noise cancellation are utilized for preprocessing EEG to reduce the a-wave and for comparison with the original EEG. A novel template and correlation coefficient method were developed for simulating thumb movement identification.
In existing researches, the experiment time of a single trial to detect the m waves is very short, therefore, the false positive results were not discussed. If the subject blinks the eyes in a trial, the set of experiment data is discarded without consideration about the false positive ratio may yield an impractical brain-machine interface. Thus in this thesis the false positive ratio would be used for evaluating the performance of a pattern recognition algorithm. Human test results indicated that the the m wave suppression phenomenon after preprocessing is more obvious than before preprocessing and the number of false positive can be reduced. On the other hand the status of subjects, i.e., nature, open eyes and close eyes, does not has effect on success ratio of a template correlation method developed in this work.
|
|---|
