It is still very difficult even for advanced simulation systems to predict how children interact with products in everyday life situations because the environment around children, including the layout of consumer products in their homes, constantly changes, and because children change their behaviors depending on the position of their center of gravity and muscular strength.
The purpose of this study is to develop a children’s climbing behavior database to clarify the configuration space of climbing behavior that enable the prediction and simulation of possible climbing postures in children.
We developed a system that collects three-dimensional posture data from color and depth images using a RGB-D camera that can capture color images and depth data, and then used pose recognition software (OpenPose) to acquire posture data. In order to construct a database of children’s climbing behavior using our developed system, we performed an observational study on how children interact with a climbing apparatus.
Fourteen children aged 20 to 58 months participated. Using depth data from Kinect (a Microsoft’s RGB-D camera) and posture data detected by Open Pose, our developed system obtains 3D coordinate data for the detected person’s posture. The coordinates of the acquired postures were converted into nine values to normalize the posture data. We purposely extracted postures only when a child touched the climbing apparatus because children react in response to the shape of an object. The extracted posture data were compressed into two or three dimensions using dimension reduction and plotted on a scatter plot. The collection of points on the plotted scatter plot can be treated as a data-driven configuration space, which indicates the space of postures that children are able to take when they climb.
We showed that it is possible to generate configuration spaces using posture data of children’s climbing behaviors.