Adaptive and Linear Energy Based Detector for a Virtual Mouse Control

The paper describes a system for controlling a virtual mouse using the mioelectric signal from voluntary contractions of the masseter and temporal muscles. The average energy of each of the data packets is compared to a threshold, established by a process of personal calibration executed before starting the system. The threshold energy value was computed using two different adaptive techniques: Linear Energy Based Detector (LED) and Adaptive Linear Energy Based Detector (ALED). Two volunteers were submitted to 90 facial contractions in order to control the mouse and test the system with the proposed techniques. The Linear Energy Based detector presented 17% of failures, the adaptive Linear Energy Based Detector presented 15% of failures and the static threshold presented 26% of failures during the commands detection. Thus the preliminary results have shown that adaptive techniques are robust alternatives for threshold events detection.