Facial Expression Recognition System: A Digital Printing Application

Human Computer Interaction (HCI), an emerging field of research in science and engineering, is aimed at providing natural ways for humans to use computers as aids. Humans prefer to interact with each other mainly through speech, but also through facial expressions and body gestures, to emphasize a certain part of that speech and display of emotions. The identity, age, gender, as well as the emotional state of a human being can be acquired from his/her faces. The impression that we receive from a reflected expression on face affects our interpretation of the spoken word and even our attitude towards the speaker himself. Although emotion recognition is seemingly an easy task for humans, it still proves to be a tough task for computers to recognize the user’s state of emotion. Progress in this field promises to equip our technical environment with means for more effective interaction with humans and hopefully, in the days ahead, the influence of facial expression on emotion recognition will grow rapidly. The application of digital printing has rapidly grown over the past few years with substantial developments in quality. Digital printing has brought about fast turnaround times and printing on demand in terms of cost. In this paper, we describe the empirical study of the state-of-the-art classifier combination approaches, namely ensemble, stacking and voting. Each of these three approaches was tested with Naïve Bayes (NB), Kernel Naïve Bayes (k-NB), Neural Network (NN), auto Multi-Layer Perceptron (auto MLP) and Decision Tree (DT).The main contributions of this paper is the enhancement of classification accuracy of the emotion recognition task on facial expressions. Our person-dependent and person-independent experiments show that using these classifier combination methodologies provide significantly better results than using individual classifiers. It has been observed from the experiments that overall voting technique with majority voting achieved best classification accuracy