Body posture measurement in a context of example-based teaching

Abstract

This paper presents a measurement process of body postures operated in a context of humanoid robot learning. The basic measured quantities are the angle joints of a human skeleton and the angle joints of a humanoid robot. Due to the differences between the two mechanical structures, the measurement results are expressed into a common representation space by the way of fuzzy scales. This paper shows how the common representation space can be defined, and presents a method to match weakly defined postures with uncertain measurements of a human posture.