Emotion detection through fusion of complementary facial features

Facial Feature extraction is used in a number of applications including emotion detection. In the following approach various popular feature descriptors, including Gabor features, HOG, DWT were computed. We have fused features using Multiview Distance Metric Learning (MDML) which utilizes complementary features of the images to extract every known detail while eliminating the redundant features. Moreover MDML maps the features extracted from the dataset to higher discriminative space. The features belonging to the same class are brought closer and those that are from different classes are forced away by the MDML thereby increasing the accuracy of the classifier employed. CK+ Dataset has been used to conduct the experiments. Experimental results represent the efficacy of the method is 93.5% displaying the potential of the recommended manner.