Methodology
The proposed method is based on a two-level CNN framework. As shown in figure 1, in order to extract emotions from an image we propose firstly background removal\cite{Kong_2016}\cite{Matsugu_2004}. We further modify the typical CNN network module to extract basic expressional vector (EV) using perceptron basic unit from a face image with background removed. The expressional vector is generated by tracking down relevant facial points of importance and EV is directly related to changes in expression.
Convolutional neural networks (CNN) is the most popular way of analyzing images. CNN is different from multilayer perceptron (MLP), such that they have hidden layers called convolutional layers. In the FERC model, we do also have a non-convolutional perceptron layer at the end. Each of the convolutional layers receives the input image, transforms it, and then outputs it to the next level. This transformation is convolution operation as given by figure 2.