Motion Analysis for Cooking Motion Recognition

Abstract

To construct a cooking motion model, we needed to analyze features of cooking motions. Our aim was to recognize cooking motions by tracking the motions of the joints of a cook's arms with a three-dimensional depth sensor. To recognize the motions, we needed to analyze their features of interest. We selected "cutting" and "mixing" as cooking motions of interest. Cutting is a motion where the cook's forearm moves up and down and the upper arm moves forward and back. Mixing is a motion where the cook's forearm moves as if drawing a wide circle in front of the body. Therefore, we focused on the motions of the forearm and upper arm as features of the cooking motion. With regard to the x-axis of the forearm, cutting had a small value for the logarithmic power, but mixing had a large value. This indicates a change in the same manner as the assumed motion. These results showed that the fifth logarithmic power of the x-axis can be used as a feature of cooking motions.