Vol.2 No.2 2009

40/98

Research paper : A secure and reliable next generation mobility (Y. Satoh et al.)−120−Synthesiology - English edition Vol.2 No.2 (2009) Question and Comment (Motoyuki Akamatsu)You mention risk detection as an important technological issue, but risk decision that matches the user’s risk recognition is extremely difficult as mentioned in section 5.1, and future R&D is essential. Therefore, the “risk detection” in section 3.5 is an explanation of the technology for obstacle detection (including dips). I think you need a description on the decision for the degree of risk and the decision to decelerate or stop.Answer (Yutaka Satoh)As you indicated, the technology for accurately detecting risks was insufficient, so we added the description in section 3.5 and newly added Fig. 8. For the analysis of level differences, we described that the detailed analysis of bumps and dips is difficult due to precision issues in the current system, and that we are separately working on a stereo image processing system using near-infrared pattern projection to solve this problem, and added a reference for this research.6 Gesture detectionQuestion and Comment (Motoyuki Akamatsu)The description about gesture detection is rather simple, and the range of application is not clear. Please provide a more detailed explanation for sensing and judgement on gestures and seating positions.Answer (Yutaka Satoh)As you mentioned, the description of gestures was insufficient so we added descriptions in section 3.6 and section 5.1. Specifically, we added the points: currently implemented are (1) the function to detect the abnormality of the seating position and (2) the function to detect the gestures of the arm; and we presupposed large motions since currently the gestures are determined by simple matching of quantized three-dimensional patterns. However, in actual fact, there is high expectation for gesture recognition from people who can move only parts of the body due to handicaps. Specifically, there is a request for recognizing the gesture of slightly moving the shoulder and we are working on it, but unlike the assumption of relatively large motion in the current technology, normal action and gesture motion cannot be separated and recognized at this point. Simple matching is insufficient, and we are considering a learning pattern-matching method. We added a description on this.

※このページを正しく表示するにはFlashPlayer9以上が必要です