Double constrained bag of words for human action recognition

作者：

Highlights：

• Construct the constraint factors of length and angle between the vectors.

• Propose length and angle constrained linear coding method.

• Generate a new presentation-level feature through hierarchical weighting.

• Propose a new training method of ELM under double constraints.

摘要

•Construct the constraint factors of length and angle between the vectors.•Propose length and angle constrained linear coding method.•Generate a new presentation-level feature through hierarchical weighting.•Propose a new training method of ELM under double constraints.

论文关键词：Human action recognition,Bag of words (boW),Feature encoding,Extreme learning machine (ELM)

论文评审过程：Received 29 October 2020, Revised 18 May 2021, Accepted 16 July 2021, Available online 4 August 2021, Version of Record 9 August 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116399