International Conference on Advances In Engineering And Technology - ICAET 2014
Author(s) : HARKISHAN SOHANPAL, RAJVIR KAUR, SONIT SINGH
In this paper, we present human activity recognition on static images. First, for feature extraction we employ Histograms of Oriented Gradients (HOG). The HOG is invariant to geometric transformations and photometric transformation such as changes in illumination or shadowing effect. The extracted features are then classified using Back- Propagation Neural Network (BPNN) classifier. Experimental results on Images from Weizmann dataset using proposed methodology show the accuracy of 99.2%. The results show that the human activity recognition can effectively be done using HOG features and BPNN as classifier.