Facial Emotion Recognition for Adaptive Interfaces Using Wrinkles and Wavelet Network

Today, the desire to interact with intelligent machines is greater than ever. In this light, facial emotion recognition research has been gradually linked to the development of systems capable of recognizing human affective state and interpreting them to enrich the user experience. Facial expression recognition is a very active area of research; many fields use human face as source of information. In this work we present a novel method for facial expression recognition. Firstly we start by designing a template for important facial components. We are interested in extracting only areas that improve the recognition of current expression. The extraction of template was based mainly on geometric transformations and the detection was through the iris position to ensure a good result. Secondly we used the Gabor wavelet to extract wrinkles in the forehead and glabella areas. The Gabor wavelet is very robust and effective for wrinkles detection. For the other components we used the Canny filter to extract edges and enhance the recognition phase. Finally to the classification phase we used the wavelet network. Experimental results show the robustness of our method to extract face features and report an average precision rate of 81.96% for emotion recognition rate of seven basic emotions.