Realistic action recognition via sparsely-constructed Gaussian processes

作者：

Highlights：

• We propose a realistic action recognition framework based on Gaussian processes (GPs).

• An ℓ1 construction and a local approximation covariance weight updating method are proposed for GPs.

• Our GPs are demonstrated to be robust to data noise, automatically sparse and adaptive to the neighborhood.

摘要

Highlights•We propose a realistic action recognition framework based on Gaussian processes (GPs).•An ℓ1 construction and a local approximation covariance weight updating method are proposed for GPs.•Our GPs are demonstrated to be robust to data noise, automatically sparse and adaptive to the neighborhood.

论文关键词：Action recognition,Gaussian processes,ℓ1 construction,Local approximation

论文评审过程：Received 6 February 2014, Revised 31 May 2014, Accepted 4 July 2014, Available online 14 July 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.07.006