Abstract
A Brain Machine Interface is a communication systemwhich connects the human brain activity to an external devicebypassing the peripheral nervous system and muscular system. Itprovides a communication channel for the people who aresuffering with neuromuscular disorders such as amyotrophiclateral sclerosis, brain stem stroke, quadriplegics and spinal cordinjury. In this paper, a simple BMI system based on EEG signalemanated while visualizing of different colours has beenproposed. The proposed BMI uses the color visual tasks and aimsto provide a communication through brain activated controlsignal for a system from which the required task operation canbe performed to accomplish the needs of the physically retardedcommunity. The ability of an individual to control his EEGthrough the colour visualization enables him to control devices.The EEG signal is recorded from 10 voluntary healthy subjectsusing the noninvasive scalp electrodes placed over the frontal,parietal, motor cortex, temporal and occipital areas. Theobtained EEG signals were segmented and then processed usingan elliptic filter. Using spectral analysis, the alpha, beta andgamma band frequency spectrum features are obtained for eachEEG signals. The extracted features are then associated todifferent control signals and a neural network model using backpropagation algorithm has been developed. The proposedmethod can be used to translate the colour visualization signalsinto control signals and used to control the movement of a mobilerobot. The performance of the proposed algorithm has anaverage classification accuracy of 95.2%.
Keywords— Brain Machine Interface, Colour visual tasks,Neural Network.
I. INTRODUCTION
Electroencephalogram (EEG) is defined as electricalactivity recorded from the surface of scalp using electrodes[13]. EEG signals are the electrophysiological measures ofbrain function. Various actions performed such as visually,mentally and physically produce different EEG signals, whichitself becomes a unique pattern for it. Thus the difference inEEG signals on various actions helps to develop a BrainMachine Interface. Brain Machine Interface (BMI) or BrainComputer Interface (BCI) is a system that providescommunication page link between the human brain and a digitalcomputer. Brain Computer Interfaces is a machine interfacewhich is made to help people, like who is paralyzed able tocontrol a machine through thought [2]. In this system, usingbrain signal, a subject can communicate and control anyphysical activity either using a machine or without a machine.BMI system is used to help people who are affected byparalysis, quadriplegics and amyotrophic lateral sclerosis todrive computers directly by brain activity rather than byphysical means. In recent years many research have beencarried out in developing BMI systems and it is mainlyinvolved in recording an electroencephalographic (EEG)signals using surface electrodes [1]. Many significanttechnological advancement have occurred in the past decadetowards developing a BCI, such as using visual evokedpotential (VEP), slow cortical potential (SCP), P300 evokedpotential, sensorimotor activity mental tasks and multipleneuromechanisms [2], [3], [4] and [5].Very few researcheshave investigated the effect of colour visualization on theEEG signal activity [7].In this paper, a simple protocol has been proposed forvisualizing different colors namely black, cyan, green,magenta, white and yellow. These six colour were chosen asthey emanates high brain activity responses and provides gooddiscriminating features [7]. The power spectral features valuesfrom three bands such as alpha, beta and gamma band EEGsignals were used as features and a neural network modeltrained with back propagation algorithm (using adaptivelearning rate and momentum coefficient) was developed todiscriminate the visual perception of different colours.
II. FEATURE EXTRACTION
A. Protocol and Data Collection EEG brain signals are recorded using the standardequipment Mindset-24 Topographic Neuro-mappingInstrument by Nolan Computer Systems LLC along with anelectrode cap [6] and [14]. This instrument is also called as1.5 to 34 Hz data acquisition system. Ten healthy righthandedvolunteers (10 men), aged between 21 and 25, haveparticipated in this experiment. All the ten subjects have noprior experience in EEG experiments. The subjects wererequested to get seated in a silent room and also requested notto make any overt movement while performing the colourvisualization tasks. A 19 channel Electro-Cap was used forrecording the brain signals from scalp as per the 10-20 systemof electrode placement [8] and the measurements were madewith reference to electrically linked mastoids, A1 and A2. A 19channel electrode cap along with the internal 10-20 electrodepositions are shown in Figure In the experimental study subjects were asked to performsix different colour visualization tasks and their correspondingEEG signals were recorded. All the subjects were free fromillness at the time of EEG recording. Before starting the datacollection, the data collection procedures were explainedclearly for each subject. The subjects were seatedcomfortably in front of a color LCD monitor and were askedto view the displayed colors. All the ten subjects were askedto view the colour screen in a relaxed condition during thedata collection of all colour tasks. Before starting the real datacollection, a sample data collection for each colour task wasconducted to find the difficulties in performing each task fromthe subject feedback. Each subject participated in five trialsessions. Each color was displayed on the color LCD monitorfor 10 seconds and left blank for 20 seconds and this processwas repeated for 10 times during a trail. For each subject,EEG signal was recorded for 10 seconds at a samplingfrequency of 256 Hz. For each subject, five such trials wereperformed. After completing every trial, the subjects wereasked to sit in a relaxed manner for 2 minutes


Download full report
http://ieeexplore.ieeeiel5/5504579/55452...er=5545339