Feature refinement: An expression-specific feature learning and fusion method for micro-expression recognition

作者：

Highlights：

• A new three-layer feature refinement framework is proposed to recognize micro-expression.

• Salient and discriminative features with attention and proposal loss are learnt.

• The first work which evaluated on two public protocols of microexpression recognition.

• Feature refinement framework achieves considerable and competitive results comparing to the state-of-the-art algorithms.

摘要

•A new three-layer feature refinement framework is proposed to recognize micro-expression.•Salient and discriminative features with attention and proposal loss are learnt.•The first work which evaluated on two public protocols of microexpression recognition.•Feature refinement framework achieves considerable and competitive results comparing to the state-of-the-art algorithms.

论文关键词：Micro-expression recognition,Deep learning,Attention mechanism,Shared feature,Feature refinement

论文评审过程：Received 23 June 2020, Revised 25 July 2021, Accepted 21 August 2021, Available online 25 August 2021, Version of Record 17 September 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108275