A novel random forests based class incremental learning method for activity recognition

作者：

Highlights：

• A novel separating axis theorem (SAT) based splitting strategy is proposed.

• By combining SAT based splitting strategy and traditional splitting strategy, we propose a novel class incremental random forests algorithm (CIRF).

• Performance of CIRF on three public activity recognition datasets is competitive and robust.

摘要

•A novel separating axis theorem (SAT) based splitting strategy is proposed.•By combining SAT based splitting strategy and traditional splitting strategy, we propose a novel class incremental random forests algorithm (CIRF).•Performance of CIRF on three public activity recognition datasets is competitive and robust.

论文关键词：Class incremental learning,Activity recognition,Random forests

论文评审过程：Received 22 May 2017, Revised 13 December 2017, Accepted 24 January 2018, Available online 31 January 2018, Version of Record 7 February 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.01.025