Accuracy Enhancement of Action Recognition Using Parallel Processing

  • C. M. Vidhyapathi
  • B. V. Vishak
  • Alex Noel Joseph Raj
Conference paper
Part of the Lecture Notes in Electrical Engineering book series (LNEE, volume 490)


Implementation of action recognition for embedded applications is one of the prime research areas in the fields of both computer vision and embedded systems. In this paper, we propose a novel algorithm to improve the accuracy of human action recognition by implementing parallel processing and incorporating multiple neural networks working in coherence for action classification and recognition. A feature set known as Eigen joints is used to model the actions in the database. The algorithm proposes an efficient method to reduce the feature set required to recognize an action accurately based on the concept of accumulated motion energy. The paper talks about the use of Robot Operating System and its advantages for implementing parallel processing. The paper also presents a comparative study in the accuracies of action recognition between support vector machine (SVM) and Gaussian Naïve Bayes (GNB) classifiers for recognizing the actions for which the networks are trained. In this paper, we also talk about how multiple supervised learning neural networks working in coherence can detect an action whose model is not present in the database.


Accumulated motion energy Action recognition Supervised learning Support vector machine Gaussian Naïve Bayes Eigen joints 


  1. 1.
    Cheng H, Hwang J (2008) Integrated video object tracking with applications in trajectory based event detection. J Vis Commun Image Represent 22(7):673–685CrossRefGoogle Scholar
  2. 2.
    Pirsiavash H, Ramanan D (2012) Detecting activities of daily living in first-person camera view. In: Proceedings of the 2012 IEEE conference on computer vision and pattern recognition. IEEEGoogle Scholar
  3. 3.
    Papadopoulos GT, Axenopoulos A, Daras P (2014) Real time skeletal tracking based human action recognition using kinect data. In: Lecture notes in computer science, vol 8325, pp 473–483Google Scholar
  4. 4.
    Wang P, Li W, Ogunbona P, Gao Z, Zhang H (2014) Mining mid-level features for action recognition based on effective skeleton representation. eprint arXiv:1409.4014
  5. 5.
    Yang X, Tian Y (2014) Effective 3D action recognition using eigen joints. J Vis Commun Image Represent 25(1):2–11MathSciNetCrossRefGoogle Scholar
  6. 6.
    Bobic A, Davis J (2001) The recognition of human movement using temporal templates. IEEE Trans Pattern Anal Mach Intell 23(3):257–267CrossRefGoogle Scholar
  7. 7.
    Wang C, Shi XH (2015) A marker-less two-hand gesture recognition system using kinect depth camera. In: IEEE international conference on signal processing, communications and computing (ICSPCC)Google Scholar
  8. 8.
    Bian W, Tao D, Rui Y (2012) Cross-domain human recognition. IEEE Trans Syst Man Cybern (Part B) 42(2)Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  • C. M. Vidhyapathi
    • 1
  • B. V. Vishak
    • 1
  • Alex Noel Joseph Raj
    • 1
  1. 1.School of Electronics EngineeringVIT UniversityVelloreIndia

Personalised recommendations