The Microsoft Kinect motion sensing device can capture the skeletal movement of a pedestrian. A skeleton in motion draws an orbit in high dimensional pose parameter space. The Karhunen-Loéve expansion can project this high dimensional orbit into two-dimensional eigenspace. A method is proposed to differentiate pedestrians by using their respective traces in the eigenspace. This uses the nearest neighbor rule to identify pedestrians on the basis of the cosine similarity between the traces. In our experiments, this method gave a personal identification rate of 34 pedestrians.
