Investigation of Preliminary Motions from a Static State and Their Predictability

Abstract

Humans observe the actions of others and predict their movements slightly ahead of time in everyday life. Many studies have been conducted to automate such a prediction ability computationally using neural networks; however, they implicitly assumed that preliminary motions occurred before significant movements. In this study, we quantitatively investigate when and how long a preliminary motion appears in motions from static states and what kinds of motion can be predicted in principle. We consider this knowledge fundamental for movement prediction in interaction techniques. We examined preliminary motions of basic movements such as kicking and jumping, and confirmed the presence of preliminary motions by using them as inputs to a neural network. As a result, although we did not find preliminary motion for a hand-moving task, a left-right jumping task had the most preliminary motion, up to 0.4 s before the main movement.

Methods of Generating Emotional Movements and Methods of Transmitting Behavioral Intentions: A Perspective on Human-Coexistence Robots

“…Xu et al [70] investigated the movements of lower limbs (button touch, kick, rotational jump, and left-right jump) from a static state that are difficult to predict because they can be performed without a large movement of the center of gravity. They reported the existence of preliminary motion in the movements of the lower limbs and the predictability of significant motion from their observations.…”