Key Frames Extraction of Human Motion Capture Data Based on Cosine Similarity

In this paper, we propose the key frame extraction method based on cosine similarity algorithm. The method removed the joints that have less influence on the human posture. Then we regard the human motion capture data as a vector in the Euclidean space, which is used to represent the motion frame. We extract the key frames by calculating the cosine similarity between vectors. Experimental results show that our method has high compression rate and low reconstruction error in the extraction of key frames, and has great practical value.

[1]  Liu Haiyan,et al.  Key frame extraction based on Artificial Fish Swarm Algorithm and k-means , 2011, Proceedings 2011 International Conference on Transportation, Mechanical, and Electrical Engineering (TMEE).

[2]  Rui Liu,et al.  Motion Key-Frame Extraction by Using Optimized t-Stochastic Neighbor Embedding , 2015, Symmetry.

[3]  Daniel Thalmann,et al.  Key-posture extraction out of human motion data , 2001, 2001 Conference Proceedings of the 23rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[4]  Guilherme N. DeSouza,et al.  A new hierarchical particle filtering for markerless human motion capture , 2009, 2009 IEEE Workshop on Computational Intelligence for Visual Intelligence.

[5]  Tolga K. Çapin,et al.  Multiscale motion saliency for keyframe extraction from motion capture sequences , 2011, Comput. Animat. Virtual Worlds.

[6]  Feng Liu,et al.  3D motion retrieval with motion index tree , 2003, Comput. Vis. Image Underst..

[7]  Yang Ta.  Extraction of keyframe from motion capture data based on quantum-behaved particle swarm optimization , 2014 .

[8]  Ksenia Oskina,et al.  Text Classification in the Domain of Applied Linguistics as Part of a Pre-editing Module for Machine Translation Systems , 2016, SPECOM.