Imagine The Motion Classification Algorithm. Brain-computer Interface Research

With the population aging and the quantity of the disabled growing, the common channel responded to environment is sometimes useless to them. So the research on brain science has attracted more attention to satisfy the demands of the people who want to overcome their disadvantages by the brain science research and the related domains. Recently, the Brain-Computer Interface (BCI) outstands from the brain science researches since it supplies a channel between human and the environment which is different from the usual brain-nerve-muscle channels. Such equipment can help the disabled and the old who have difficulty in limb movement to obtain the ability to communicate with others again.Research on the BCI based on the electroencephalogram (EEG) gets much more attentions in BCI researches thanks to its characteristic of simplicity, safety and noninvasive. The research done in this thesis is to classify the EEG which is recorded when a person imagines the movement of left hand or right hand in a BCI.On the base of former studies, a BCI frame of left/right hand movement imagination (MI) is designed with the introduction, research and innovation to every part of a BCI system in this thesis. The main works of this thesis are as follows:1) A new experiment is designed to record the EEG according to the requirement of the subject on the basis of the EEG record equipment and the former softwares in the laboratory. And then a software named MIEEG is devised using VC++6.0 to meet the requirements of the experiment. The software MIEEG designed can be used as a part of the SDUND which is also a software designed in my laboratory before. Besides it can also be used as an independent EEG recording software to record EEG during the MI. This software can be extended easily for more research in the future. 2) The position used to record EEG is confirmed by reference of the brain function subareas and by choosing suitable electrode and lead-mode. After filtering the recorded EEG data in frequency domain, the sensitive band for MI is selected. Then handling of normalization on the data is used to minimize the error. The sensitive time band for MI in time domain is also selected in which the EEG features are extracted and quantized. The feature used in this thesis is Event-Related Synchronization/Desynchronization (ERD/ERS) which is sensitive to imagery movement. Two feature values including the effective field M and the space complexity Ω are quantized for the ERD/ERS. To get a high accuracy, the subtraction is handled on the quantized features Ω to make it more favorable for classification.3) Two classifiers containing Fisher linear discrimination function and Support Vector Machine (SVM) are introduced in this thesis. Their construction principles and formula derivations are also explained .The main task in this step focus on how to construct the SVM, select the core function and the parameters. When the two features are input to the SVM, artificial nerve network is presented in this thesis to simulate the human thoughts and a higher accuracy is achieved.The dataset of 2003 international BCI competition are used for research purpose, and dataset recorded by MIEEG are adopted for testing the algorithm. The classification accuracy using SVM on features extracted from EEG can reach 87.1% and 82.2% while using Fisher linear discrimination function can reach 81.2% and 76.6%. The results show that SVM performs better than Fisher linear discrimination function during left/right hand motor imagery task.