Gestures Classification Based on Semantic Classification Tree

This paper presents a novel approach for classifying arm gestures by building a three-layer semantic classification tree. We first extract three semantic features that explicitly contain communicative information and have outstanding descriptive ability. Based on the analysis of human body parts and the spatial-temporal characteristic of gestures, collaboration between two arms, key posture, as well as the motion periodicity are extracted. These features function as a good way of building bridges between bottom-layer features and high-layer semantics. We then construct a classification tree that incorporates the semantic features on different layers by prior knowledge. In particular, two different classifiers, i.e. GentleBoost and NearestNeighbor, are employed on different layers. The simplicity and efficiency of our method lies in that, the multi-layer hierarchy makes it possible to use very simple features on each layer, moreover, based on the motion attributes of human gestures, we arrange the structure of our classification tree manually, which makes our classification more reliable.