الفهرس | Only 14 pages are availabe for public view |
Abstract Brain Computer Interface (BCI) is a hot research area that has grown in rehabilitation, biomedical and electrical engineering, computer sciences and Virtual Reality fields. BCI aims to enhance the quality of life for all humans. Spelling is one of the challenges BCI applications, which allow people to type numbers, characters, words, or sentences by recording the users’ brain activity. In this thesis, we examine the performance of mental speller that detects the number thought of by the subjects. Electroencephalograms (EEG) signal that records the brain activities are acquisition using the Emotive EEG headset with fourteen channels. The channels, distributed according to international 10-20 system. The recording signals are preprocessed using Independent Component Analysis (ICA), and Auto Regressive (AR) as features extraction, finally utilized Support Vector Machine (SVM) and K Nearest Neighbors (KNN) for various Ks as classification phase. Results were obtained using the Matlab software, which is considered one of the most powerful mathematical software tools. The results show that using SVM classifier has achieved an average Correct Classification Rate (CCR) of only 11.6%. KNN on the other hand has achieved a CCR of 85.1% when only the vote from one neighbor is considered. Increasing the number of voters positively affects the average results. This can be seen when k increased to 3, 5, 7, and 9. Their classification results moves to 87.3%, 90%, 92.8%, and 94.3% respectively.The results have shown the efficiency of mental speller with imagined writing activity. The proposed system has achieved good results in both performance accuracy and character recognition time. The accuracy results get better when higher values for K are deployed. Since the proposed system is to give a new way of communication channel, we hope that this project will be useful for disabled to help them spend their life as good as others will. |