Speeded Human Action Recognition Based on Action Snippets Algorithm and Neighborhood Component Analysis
Sanjay T.Gandhe1, Pravin A.Dhulekar2
1Dr Sanjay T. Gandhe, Professor, Department of Electronics and Telecommunication Engineering, Sandip Institute of Technology and Research Centre, Nashik, Savitribai Phule Pune University, Pune, (Maharashtra), India.
2Pravin A. Dhulekar, Research Scholar, Department of Electronics and Telecommunication Engineering, Sandip Institute of Technology and Research Centre, Nashik, Savitribai Phule Pune University, Pune, (Maharashtra), India.
Manuscript received on 02 June 2019 | Revised Manuscript received on 10 June 2019 | Manuscript published on 30 June 2019 | PP: 2609-2615 | Volume-8 Issue-8, June 2019 | Retrieval Number: H7450068819/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The human action recognition has large number of applications in various domains however its applicability in many real time scenarios has been restricted due to the speed of action recognition algorithms. Therefore, there is great need to develop the faster action recognition algorithm. To address this constraint we present a hybrid algorithm for speeded human action recognition (SHAR) based on action snippets algorithm and neighborhood component analysis. The action snippet algorithm results in filtering the key frames and eliminating less informative frames. Moreover, the feature selection algorithm at further stage of recognition helps to trimmed down the length of feature vector massively. These two algorithms are integrated with standard pair of feature extraction and classification at different stages to accomplish the goal of faster recognition without compromising the recognition accuracy. The performance of proposed algorithm has been evaluated on weizmann dataset. The proposed methodology provides overall recognition accuracy of 90% with four times faster classification over standard approach of feature extraction followed by classification.
Keyword: speeded human action recognition (SHAR), action snippets algorithm and neighborhood component analysis.
Scope of the Article: Component-Based Software Engineering.