In the last years, gesture recognition has gained increased attention in Human-Computer Interaction community. However, gesture segmentation, which is one of the most challenging tasks in gesture recognition applications, is still an open issue. Gesture segmentation has two main objectives: first, detecting when a gesture begins and ends; second, recognizing whether a gesture is meant to be meaningful for the machine or is a non-command gesture (such as gesticulation). This paper proposes a novel test protocol for the evaluation of different techniques separating command gestures from non-command gestures. Finally, we show how we adapted adopted our test protocol to design a touchless, always available interaction system, in which the user communicates directly with the computer through a wearable and “intimate” interface based on electromyographic signals.