Action recognition via structured codebook construction
作者：
Highlights：
• We propose a novel structured codebook construction method for action recognition.
• Structured codebook encodes rich spatial and temporal contextual information.
• The structured codebook is built by exploring statistical properties of elementary actions.
• The novel representation is robust to unwanted background features.
• The proposed method prompts the recognition accuracy.
摘要
Highlights•We propose a novel structured codebook construction method for action recognition.•Structured codebook encodes rich spatial and temporal contextual information.•The structured codebook is built by exploring statistical properties of elementary actions.•The novel representation is robust to unwanted background features.•The proposed method prompts the recognition accuracy.
论文关键词：Action recognition,Bag-of-words models,Structured codebook,Sparse coding,Contextual information
论文评审过程：Received 4 May 2013, Revised 18 October 2013, Accepted 16 January 2014, Available online 15 February 2014.
论文官网地址：https://doi.org/10.1016/j.image.2014.01.012
