Fast and Accurate Smartglass Angles Inference Based on Periodic Behavior in Walking

Smartglasses are expected to be applied to various consumer applications, such as pedestrian navigation systems, games, and business support applications. In such applications and systems, it is important to infer the smartglass angles while the user is in motion, especially when the user is walking, in order to understand the user's current state and motion. In this paper, we propose a method for inferring the smartglass angles while the user is walking. The proposed method focuses on the periodic behavior during the user's walking and effectively identifies the vertical downward direction of the smartglass. Then it infers the smartglass angles, i.e., which direction the smartglass heads to, fast and accurately. Through the experimental evaluations, the inference accuracy of the proposed method is almost the same as the baseline method, which is known as accurate enough, and the time required for convergence is reduced to 42.1% for the roll angle inference during walking, and 54.4% for the pitch angle inference during walking.