Electroencephalogram signal interpretation system for mobile robot

• Main Author: Hasan, Intan Helina
• Format: Thesis
• Language: English
• Published: 2013
• Online Access: http://psasir.upm.edu.my/id/eprint/67598/1/ITMA%202013%208%20IR.pdf; http://psasir.upm.edu.my/id/eprint/67598/

In recent years, Brain-Computer Interface (BCI) research has provoked an enormous interest among researchers from different fields since it is an important element in assistive technology. The most popular approach is a non-invasive method, using Electroencephalogram (EEG) analysis which acquires signals from the brain. Currently, the BCI application is to acquire signals from 32 to 64 electrodes’ recordings and translate them to a movement using various computing algorithm which can be used in wheelchair navigation, or control robot movements. However, it will be time consuming and an exhausting experience if the single command translation from large number of electrodes is used to help physically disabled and elderly people with their daily tasks or chores. An improved interface needed to be developed to allow BCI to become a user-friendly interface for the targeted groups. The aim of this project is to develop an algorithm that can choose optimal four electrodes for signal recording, and convert one thought into multiple commands with the chosen electrodes. Using sample datasets, the EEG signal is analyzed to determine the most suitable scalp area for P300 detection, while optimization with genetic algorithm (GA) is developed to select best four channels. Next, a signal interpretation system is designed and developed to translate the signal and send the pre-programmed commands to the robot through the operating computer. Based on the analysis and optimization of the datasets, P300 signals are most clear and robust at the midline and parietal area of the scalp, and can be detected at around 500ms after a stimulus. After 30 GA runs, the optimal four sets of electrodes are chosen based on their coefficient of determination or r² values, where higher values contributes to higher repetition rates. Using signals from the chosen four electrodes to evaluate the signal interpretation system, a success rate of 75-80% is received. With this system, user can expect a more convenient preparation with lesser electrodes used, and faster execution of the robot commands since they are pre-programmed according to user’s intention and selected route.