Real-world facial expression recognition using metric learning method

Abstract

Real-world human facial expressions recognition has great value in Human-Computer Interaction. Currently facial expression recognition methods perform quite poor in real-world compared with in traditional laboratory conditions. A key factor is the lack of reliable large real-world facial expression database. In this paper, a large and reliable real-world facial expression database and a Modified Metric Learning Method based on NCM classifier (PR-NCMML) to regress the probability distribution of emotional labels will be introduced. According to experiments, the six-dimension emotion probability vector derived by PRNCMML is closer to human perception, which leads to better accuracy than the state-of-the-artmethods, such as theSVMbased algorithms, both dominant emotion prediction and multi-label emotion recognition.