On recognition of human orientation with respect to a static camera

Abstract

Detection of orientation of a human being with respect to a static camera is an important problem in computer vision. This paper proposes a method for detection of orientation of a human being from four different orientations (e.g. front, side-left, side-right, back) with respect to a static camera. In the proposed method, extraction of features of a human being has been performed in terms of boundary description known as the signature. A template database of four different human orientations has been created. The extracted features of a testing sample have been compared with that stored in the template database and a dissimilarity value has been calculated. The classification has been performed using the dissimilarity value as a metric. The results obtained are encouraging.