I. Introduction
Recently, wearable devices are expected to realize an ubiquitous computing environment in the near future. In this paper, we consider an input device using a wearable camera and a human finger. The target video is assumed to be captured by a wearable camera that is equipped in grasses for example. Then, the video sequence is processed in order to capture the movement of the finger, e.g., a click, or a swipe. For that purpose, we consider using a foreground segmentation method to accurately obtain the move of a finger and an arm.