Analysis And Design Of EEG Based Brain Computer Interface Remote Control System

Electroencephalography is the neurophysiologic measurement of the electrical activity of the brain by recording from electrodes placed on the scalp. EEG consists of potentials generated by neuron cells on cortex, and is a noninvasive and high temporal-resolution technique which can directly reflect brain activity. EEG based Brain Computer Interface (BCI) offers new communication and control channels between human brain and computer that does not require periphery neurons or muscle control. While BCI technology originally aimed to establish a direct communication channel between brain and computer for paralyzed people, recent BCI research has expanded to a broader area that may benenifits normal people as well, for instances, remote system controlling, virtual environment exploring, and new generation game controlling. Meanwhile, among available techniques for brain signal acquisition, EEG is preferred for real time BCI systems due to its high temporal resolution, portable size, noninvasive recording and lower cost.This thesis first introduces the background research work that related to EEG based BCI systems and then propose a framework of BCI system, consisting of several modules, which are EEG acquisition, signal preprocess, feature extractor, pattern classifier, remote viehicle control and data visualization system.Our BCI system is developed to provide a general platform for research and BCI system design. With an extendable architecture, our system integrates alternative feature extraction methods, classification algorithms, control tasks in different user populations and real-time capability, supporting more fiexible experiment schemes and scalability. The system is able to adapt to subjects online and to work in real time condition with high accuracy and low latency. The BCI system is based on analysis of EEG patterns of objects'left and right motor imagery. Independent Component Analysis and temporal filter are employed for artifacts removal and noise reduction. Wavelet transformation, common spatial pattern analysis and common discriminant tensor analysis are used as feature extractor, from which feature vectors are classified by SVM.In the training stage, four subject's EEG signals were recorded and analyzed.We use different trail lengths to approach the best model for specific subject. After a proper time of self-adaption and training, they ware required to drive the vehicle from one point to another. Such real navigation could have favorable impact on EEG dynamics, making the whole procedure easier for general subjects. The successful experiment had proved that three subjects were able to control the remote vehicle smoothly.The best classification accuracy of all objects' is up to 99% in offline analysis and 95% in online condition. So we can draw the conclusion that the developed BCI system was successfully applied to control remote vehicle.