Motion descriptors for micro-expression recognition
作者：
Highlights：
•
• A fusion motion boundary histograms (FMBH) is proposed. The FMBH establishes the connection between two feature components in motion boundary histograms feature vector, through a non-linear mapping. Experimental results indeed prove that such a connection improves the recognition accurate ratio.
• In contrast with current studies that compute the histograms of motion features by exploiting a fixed number of spatial orientation bins, this study investigates the influence of the number of orientation bins and its rotation so as to test the corresponding recognition performance.
摘要
•A fusion motion boundary histograms (FMBH) is proposed. The FMBH establishes the connection between two feature components in motion boundary histograms feature vector, through a non-linear mapping. Experimental results indeed prove that such a connection improves the recognition accurate ratio.•In contrast with current studies that compute the histograms of motion features by exploiting a fixed number of spatial orientation bins, this study investigates the influence of the number of orientation bins and its rotation so as to test the corresponding recognition performance.
论文关键词：Micro-expression recognition,Optical flow,HOF,MBH,FMBH
论文评审过程：Received 26 September 2017, Revised 28 April 2018, Accepted 21 May 2018, Available online 24 May 2018, Version of Record 21 June 2018.
论文官网地址：https://doi.org/10.1016/j.image.2018.05.014
