Facial Expression Recognition Based on Adaptive Robust Online Metric Learning Algorithm
Zhu Erl · Video Engineering · 2015
For the problem ofnoise tag in spontaneous facial expression recognition,an adaptive robust online metric learning method is proposed.Firstly,a new metric space is learned to increase the discrimination of different facial expressions.Then,the sensitivity and specificity is defined to represent each annotator.The latent variables representing real class label are introduced,and iteration is used to solve the distance measure and the reliability of the annotator in expectation maximization architecture.Experimental results on MFP and AR face databases show that proposed method has higher recognition accuracy than several other advanced methods,its accuracy on happy expression can achieve 99.7%,and it has partly reduced the computation overhead.