Robust basketball sports recognition by leveraging motion block estimation

作者：

Highlights：

• We fuse motion features, posture information and skeleton sequence to describe human actions.

• We first select motion regions based on large-scale affinity graph construction, where the entropy of the selected motion regions is calculated for extracting the motion block.

摘要

•We fuse motion features, posture information and skeleton sequence to describe human actions.•We first select motion regions based on large-scale affinity graph construction, where the entropy of the selected motion regions is calculated for extracting the motion block.

论文关键词：Human action recognition,Human–computer interaction,Sports narrative,Motion block estimation

论文评审过程：Received 10 September 2019, Revised 23 December 2019, Accepted 10 January 2020, Available online 16 January 2020, Version of Record 20 February 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.115784