Local part model for action recognition

作者：

Highlights：

• We propose a new local part model for action recognition.

• A feature sampling strategy with high feature density is used.

• We explore and prove the benefits of using accurate optical flow algorithm for action recognition.

• High performance and fast action recognition are achieved.

摘要

•We propose a new local part model for action recognition.•A feature sampling strategy with high feature density is used.•We explore and prove the benefits of using accurate optical flow algorithm for action recognition.•High performance and fast action recognition are achieved.

论文关键词：Bag-of-features (BoF),Action recognition,Random sampling,Local part model,Multi-channel SVM

论文评审过程：Received 26 June 2014, Revised 10 August 2015, Accepted 18 November 2015, Available online 16 January 2016, Version of Record 31 January 2016.

论文官网地址：https://doi.org/10.1016/j.imavis.2015.11.010