Proceedings of the 2008 IEEE, CIBEC'08 978-1-4244-2695-9/08/$25.00 ©2008 IEEE CLASSIFICATION OF THE IMAGINATION OF THE LEFT AND RIGHT HAND MOVEMENTS USING EEG M. A. Hassan 1 , A. F. Ali 1 , M. I. Eladawy 2 1 Department of Biomedical Engineering, Helwan University, Helwan, Cairo, Egypt 2 Department of Communication and Electronics, Helwan University, Helwan, Cairo, Egypt E-mail: Biomedo@yahoo.com Abstract-Brain-computer interface (BCI) is a new and promising area of research which is assumed to help in solving a lot of problems especially for handicapped people. Detection of the imagination of the left and right hand movements can be used to control a wheelchair accordingly. Fortunately, modification of the brain activity caused by the imagination of the left or right hand movements is similar to the modification observed from a real left or right hand movements. The electrical activity of these modifications can be picked up from scalp electroencephalogram electrodes. In this work, we introduce a new method to detect and classify the imagination of the left and/or right hand movements. This method is based on exploring the time domain information in both alpha and beta rhythms using complex Morlet wavelet transform. Then, the fast Fourier transform is applied to explore the frequency domain information. The extracted features using both time and frequency domain information are then reduced using a feature subset selection algorithm. Then, the reduced features were fed into a multilayer backpropagation neural network to classify left from right hand movement imagination. The experimental results showed that the algorithm has reveals classification accuracy rates ranges from 97.77% to 100%, which are superior to the classification accuracy rates compared to other techniques. Keywords - brain computer interface, motor imagery, feature subset selection, EEG classification I. INTRODUCTION The human brain is the control unit that controls all functions the human can perform such as vision, hearing, moving…etc. However, the control commands are sent as neural signals originated from its dedicated mesial structures in its functional area, projecting into the motor cortex, located over the lateral surface of the brain [1]. From there, the neural commands are relayed through descending tracts to the spin, with most of the descending nerve fibers developing a cross over to a contralateral side. In the spin the appropriate motor groups of the destination organ are simulated, consequently the destination organ performs command action. Additionally, the neural signals of Right and left hand movements are controlled by the sensorimotor areas in the brain like all muscle movements in the human body. Fortunately, The imagination of (i.e. the concentration in) movement activates the same brain areas/functions that can be activated in programming and preparing to a real movement. Thus, neural signals of motor imagination are similar to neural signals of real movement but blocked at some corticospinal levels. Currently, a lot of techniques can monitor brain activity. These include, for example, functional Magnetic Resonance Imaging (fMRI) [2], magnetoencephalography (MEG) [3], Positron Emission Tomography (PET), Single Photon Emission Computer Tomography (SPECT) [4], single neuron recording (with microelectrodes), and electroencephalography (EEG) [5]. Although, fMRI, PET, and SPECT are more accurate, and have more spatial resolution they are not candidate for BCI applications due to its main characteristics as large device that are heavy weight and can not act as a portable device. Furthermore, single neuron recording requires that the electrodes are inserted inside the skull. Therefore, only Electroencephalogram (EEG) is candidate to monitor the electrical brain activity due to it has the following characteristics: • Better temporal resolution. • Portable. • Cost effective comparable with other brain imaging techniques. EEG picks up electrical brain activity from different brain areas using a set of electrodes attached to the skull and distributed according to certain configuration. However, this activity is oscillatory in nature due to the high population of neurons which form a highly complex network with feedback loops [6]. Furthermore, different six oscillations [7] are known to appear in the electrical brain signals, their types, frequencies and associated actions are listed in table I. When EEG signal is recorded from sensorimotor area; the alpha rhythm is called mu rhythm. From table I, one can notice that the oscillations in both alpha (mu) and beta rhythms are the most obvious indicators of movement. Moreover, Motor imagination of either right or left hand movement causes an event related to desynchronization at contra-lateral hemisphere sensorimotor representation area (i.e. attenuation of the amplitude of the beta and mu rhythms), and causes an event related synchronization at ipsi-lateral hemisphere sensorimotor representation area (i.e. amplification of the amplitude of beta and mu rhythms) [8,9]. In this research, EEG signals recorded from C3 & C4 electrode positions using the 10- 20 international electrode configuration system [10] are analyzed to monitor the electrical activity of sensorimotor areas of left and right brain hemispheres through extracting a set of discriminative features that enable us to differentiate between left and right hand movement imagination. Then, the extracted features were fed into an appropriate classifier to differentiate between left hand movement imagination and right hand movement imagination. TABLE I ELECTRICAL BRAIN WAVES rhythms / band Frequency Range [Hz] Happened during Delta 0.5 – 3 Deep dreamless sleep Theta 4 – 7 Drowsiness, mental imagery Alpha (mu) 8 – 12 Relaxation, and Sensory, and Motor activity Βeta 12 – 30 Active concentration, and Motor idling Gamma 26–100 certain cognitive or motor functions