Abstract
Electroencephalography (EEG) signals recorded from the Motor Cortex and Electromyography (EMG) signals recorded from the facial muscles are used to control a prosthetic hand. Preprocessing and pattern recognition phase of EEG signal analysis is performed by the Emotiv Software Suite. The proposed Brain Computer Interface System is using seven inputs to control a motorized prosthesis. These inputs range from EEG recorded from mental tasks to EEG and EMG recorded from actual limb movements. Imagined limb movements will be detected based on the brain wave rhythm at specific locations on the scalp. Each type of detected EEG pattern will be used to control a specific finger on the prosthetic hand. Electromyography signals will be recorded from eye movement and eye brow movement. Giving the user the ability to manipulate each finger individually will allow the user to perform tasks that are not possible when using body powered prosthetic hand or the EMG based prosthetic hand. The accuracy of individual finger manipulation was further increased by combining two different EEG/EMG inputs to control one desired output. Based on the user's EEG and EMG signals the following were most accurately detected by the Emotiv Suite: left/right eye wink, imagined left hand movement, clenching of team, smirk, and raising of both eyebrows. While testing for EEG/EMG compatibility, the following combinations were found impossible to be used at the same time: clenching + raising eyebrows, clenching + imagined hand movement, and winking with both eyes. The following combination of EEG/EMG signals showed good results: left eye wink + imagined left hand movement, right eye wink + imagined left hand movement, clenching + left eye wink, clenching + right eye wink, clenching + look right, raising eyebrows + look right, and smirk left + right eye wink. Hand grabber from Toysmith was redesigned to fit the needs of this project. Arduino Uno R3 Board was used to control five 180 degree servo motors. Emokey software was used to send the detected EEG patterns from the Emotiv Software to the Arduino serial monitor window. One possible future study would be to use a 32 or 64 channel EEG headset to redo this research project and compare the accuracy of controlling each individual finger. The 32/64 electrode EEG headset would help in detecting multiple imagined tasks more accurately, thus giving us additional input commands to be used for controlling the prosthesis.