能動的振動入力による手形状識別

抄録

In this paper, we propose a novel method for identifying hand poses by emitting vibrations actively, toward intuitive user interfaces. The proposed method is an active acoustic sensing method, which can simultaneously identify hand poses and provide haptic feedback. We used the power spectrum of the propagated vibration as a feature to identify when the hand pose changed, and data compression was applied to improve the precision of hand pose identification, using a support vector machine. We evaluated the recognition rates, which depend on the number of estimation targets and the compression ratio. Further, we confirmed the feasibility of our proposed method through evaluation experiments. It was found that the recognition rate of seven hand poses was around 90%, and there is little reduction in the recognition rate when haptic feedback is provided.