Detection of distinguishing features using selection methods for robot-assisted rehabilitation system, rehabroby

Task involvement is a key factor in sustaining subjects' participation in rehabilitation programs. An appropriate challenging rehabilitation task can increase engagement of the subjects. In this way, it is desirable that task difficulty must be suitably challenging to acquire great performance from rehabilitation tasks. In order to find the appropriate challenging level for each subject, it is important to detect the subject's feelings (he/she is either getting to be noticeably exhausted or disappointed), and after that to change the rehabilitation task to better suit the subjects capacities by considering their feelings. In this thesis, three main biofeedback sensors Blood Volume Pulse (BVP), Skin Conductance (SC), and Skin Temperature (ST) are used to detect the feelings of the subjects when they use a robot-assisted rehabilitation system called RehabRoby. It is also important to know which features are distinctive to properly detect the feelings of the subjects from the physiological signals acquired by these biofeedback sensors. In this thesis, we explore the distinctive features from physiological signals using both sequential forward selection (SFS) and ANOVA feature extraction methods.