Performance Comparison of Electromyogram-based Computer Cursor Control Systems

The intent of this paper is to present performance evaluations for two systems that utilize electromyogram (EMG) signals from muscles in the face of the user to control the computer cursor. Both systems translate facial muscle contractions into five cursor actions: left, right, up, down and left-click. The systems perform this translation by using features derived from the spectral analysis of the EMG signals captured. The more recent system is a modification of the original system presented in [2, 3], and has a new electrode set-up and translation algorithm. An experiment was created to record the point-and-click performance of both systems when being used by various subjects. The resulting point-and-click trial times were further analyzed using a t-statistic and Fitts’ Law. Also included are results from Matlab simulations comparing both systems. These results indicate that the newer, modified system provides more accurate classifications of the muscle contractions and faster performance when compared to the original system for the population of users involved. Key-Words: Electromyogram, EMG, human, computer, interface, spectral, analysis