ISSC 2009, UCD, June 10-11th A Novel Paradigm for Multiple Target Selection Using a Two Class Brain Computer Interface Vaibhav Gandhi *,** , Girijesh Prasad * , Damien Coyle * , Laxmidhar Behera * and Martin Mc Ginnity * *Intelligent Systems Research Centre University of Ulster **email: gandhi-v@email.ulster.ac.uk  Abstract – A brain computer interface (BCI) allows a person to communicate with external devices using electroencephalogram (EEG) or other brain signals. A typical BCI scheme consists of data acquisition, feature extraction and classification. Using the classifier output, a control command is issued to the intended devices and the subject is provided appropriate feedback. As a part of feedback, a graphical user interface (GUI) plays a very important role as a front-end display for the BCI user and enhancing overall communication bandwidth. This paper focuses on the interface design aspect of a BCI so as to provide effective control of a wheelchair or robot arm application. A motor imagery prediction based paradigm is used to create a semi synchronous interface with a focus on presentation of a new task for selection as well as to optimally utilize the subject intentions. From a theoretical assessment, it is expected that the overall time required to select from six choices using the proposed GUI will be much less compared to existing designs. Also, being a two class paradigm, it is expected that the probability of error occurrence is minimized along with a quicker traverse between choices and this may allow a limited bandwidth BCI to operate an external device with multiple degrees of freedom and choose from multiple different choices efficiently and effectively. Keywords  interface, motor imagery, multiple target, wheelchair/robot.  I. INTRODUCTION A Brain Computer Interface (BCI) is a means to provide communication between the human brain and a computer. The most common purpose of a BCI is to enable a highly disabled person to have effective control over devices by using brain waves/EEG signals (which represent the intent of the person). As shown in Figure 1, BCI involves sub tasks such as signal acquisition, pre-processing, feature extraction, classification and feedback and device commands. Figure 1 : Basic design and operation of any BCI system. Signal processing forms the basis of any BCI system since it contributes to extracting meaningful information from the brain signal, and therefore has a substantial effect on the classification outcome, accuracy and usability of the BCI. The classification of the signal (after pre processing and suitable feature extraction) represents a means to find a rule which labels an input to one of the designated output classes. However, there is a need for innovative GUI to enhance the limited bandwidth of BCI. This paper focuses on this aspect of BCI. A BCI can be based on one or more of the following cognitive events or processes:- motor imagery (MI), P300, Visually Evoked Potential (VEP), slow cortical potentials (SCP) or activity of a single neuron (invasive). P300 (an event related potential) [1] [2] is evoked in EEG as a positive peak at about 300 ms after a stimulus is focused on whereas a VEP is generated in response to visual stimulus such as flashing lights [3] at about 100 ms after the stimulus [4]. SCPs reflect shifts of potential in the cerebral cortex and normally last for about 0.5-10s [4]. The amplitude of these potential shifts varies within a range and reaches a maximum at the Signal Acquisition Pre processing Feature Extraction Classification Feedback