Recognition of student emotion based on matrix-1 median fisher's face and backpropagation algorithm

Emotions drive learning success because they hold a willingness to process information. However, it is a challenge for understanding the emotions of student in the real class. In this study, we proposed recognition of student emotion using matrix-1 median fisher's face and backpropagation algorithm. The computation of backpropagation is influenced by neuron architecture which is can be handled by feature reducing, such as fisher face. However the number of fisher's vector due to the number of class. In order to map the lower dimensional feature space than fisher's face vector, we proposed matrix-1 median of the fisher's face. In this proposed method, after face is detected, LDA on PCA space is employed for getting the fisher's face. Then fisher face is transformed into fisher's median. The backpropagation algorithm is trained using this feature to distinguish student emotions. The performances of proposed algorithm are evaluated on the UM's learning video using accuracy and iteration consuming. Our proposed method reach accuration of overly interest, interest and bored up to 0.83, 0.91, and 1, whereas original fisher face reach accuration of overly interest, interest and bored up to 0.83, 0.91, and 0.91. Combination of backpropagation and matrix-1 median fisher face need 9 iteration for training. Whereas the combination of backpropagation and fisher Face need 11 iteration. Experiment result shows that our proposed method outperform than the existing method.