Quantifying Micro-expressions with Constraint Local Model and Local Binary Pattern

Micro-expression may reveal genuine emotions that people try to conceal. However, it’s difficult to measure it. We selected two feature extraction methods to analyze micro-expressions by assessing the dynamic information. The Constraint Local Model (CLM) algorithm is employed to detect faces and track feature points. Based on these points, the ROIs (Regions of Interest) on the face are drawn for further analysis. In addition, Local Binary Pattern (LBP) algorithm is employed to extract texture information from the ROIs and measure the differences between frames. The results from the proposed methods are compared with manual coding. These two proposed methods show good performance, with sensitivity and reliability. This is a pilot study on quantifying micro-expression movement for psychological research purpose. These methods would assist behavior researchers in measuring facial movements on various facets and at a deeper level.