Violence Recognition Using Harmonic Mean of Distances and Relational Velocity with K-Nearest Neighbour Classifier
Violence recognition falls in the domain of action recognition which has gained considerable attention and importance due to its wide application. Violence recognition has to take place in real time. One main approach to accelerate the recognition is to efficiently choose and calculate suitable features to be used in recognition which is known as feature selection. This paper proposes the use of only nine harmonic means of relational distances between pairs of six joints and one relational velocity between 2 joints. The selected joints are chosen carefully based on having the highest information gain for the recognition. The results show that very high accuracy rate of 99.8 % can be achieved with k-nearest neighbours (k-NN) classifier. This excellent recognition rate would encourage researchers in trying to implement the proposed approach in hardware, as it uses comparatively few data for processing with simple algorithms.
KeywordsViolence recognition, human action recognition, feature extraction Skeleton, harmonic means, velocity, classification
The authors of this paper like to thank Yun et al.  for their generosity in sharing the dataset for use in this work.
- 2.Li, W., Zhang, Z., Liu, Z.: Action recognition based on a bag of 3d points. In: Computer Vision and Pattern Recognition Workshops (CVPRW), IEEE Computer Society Conference, pp. 9–14, California (2010)Google Scholar
- 4.Masood, S.Z., Ellis, C., Nagaraja, A., Tappen, M. F., LaViola Jr., J.J., Sukthankar, R.: Measuring and reducing observational latency when recognizing actions. In: Computer Vision Workshops (ICCV Workshops), 2011 IEEE International Conference, pp. 422–429. IEEE, Barcelona (2011)Google Scholar
- 5.Sung, J., Ponce, C., Selman, B., Saxena, A.: Human activity detection from RGBD images. Plan Act. Intent Recognit. 64, 47–55 (2011)Google Scholar
- 6.Yun, K., Honorio, J., Chattopadhyay, D., Berg, T.L., Samaras, D.: Two-person interaction detection using body-pose features and multiple instance learning. In: Computer Vision and Pattern Recognition Workshops (CVPRW), 2012 IEEE Computer Society Conference, pp. 28–35. Rhode Island (2012)Google Scholar
- 7.Liu, T., Song, Y., Gu, Y., Li, A.: Human action recognition based on depth images from microsoft kinect. In: 2013 Fourth Global Congress on Intelligent Systems (GCIS), pp. 200–204. IEEE, Cape Town (2013)Google Scholar