Abstract:
A Brain Computer Interface (BCI), sometimes called a Brain Machine Interface (BMI) is a communication device between the brain and an external device, usually a computer. The main purpose of BCI systems is repairing or assisting human motorsensory functions by asking the brain to control synthetic devices, computer cursors or robot arms. In order to extract information from the brain, physical source of information must be selected rst. Electroencephalography (EEG), Magnetoencephalography (MEG) and Functional Magnetic Resonance Imaging (fMRI) could be the sources of information. In this thesis, both acquisition hardware and software of a two channel EEG based brain computer interface was designed. EEG based BCI systems are usually implemented by analysis and classi cation of speci c features or patterns in the spontaneous or event related EEG activity. After investigation of the components in EEG, motor imagery related mu and beta rhythms were selected for the information sources of the system. In order to discriminate left and right hand movement imagery, three di erent feature extraction methods were developed using: Discrete wavelet transform, power spectrum transform and band pass FIR lters for Mu and Beta rhythms. These features were used as inputs to a two layer feed forward back propagation neural network for classi cation. Designed system was trained and simulated with the data provided in BCI Competition II. With the direction of the results, a low power system with the TI MSP430 microcontroller using FIR lters and a neural network was implemented.|Keywords: Brain Computer Interface, Motor Imagery, EEG Feature Classication, BCI.
