1Department of Computer Science and Engineering, Thiagarajar College of Engineering, Madurai, Tamil Nadu, India
2Assistant Professor, Department of Computer Science and Engineering, Bannari Amman Institute of Technology, Sathyamangalam, Tamil Nadu, India
New ideas in motion capture and virtual reality applications have been made possible by the recent advancements in motion recognition. Using a pattern recognition algorithm, Microsoft kinect can detect the different body postures of the human body. But due to the single-depth camera, the accuracy of recognition is not significant. This also has a significant effect due to the surrounding environment. The accuracy is still reduced, when the body parts are not recognized correctly. In this paper we propose a method to remove noise and increase the efficiency of the posture recognition. A similar set of postures are synthesized based on the position of the body parts. These values are then learned by a machine to give the posture of the human. This is plied to the frame based optimization to create a new posture which resembles the older one.
Posture reconstruction, local principal component analysis, Human computer Interaction
