Recognize Facial Expression Using Active Appearance Model and Neural Network

We present an image processing pipeline to recognize facial expression by first using a face template to identify a set of feature points on faces and then applying a neural network to classify facial expression to one of six categories, namely, happy, surprise, sad, distracted, focused, and plain. We tested the pipeline on standard database and found that it can achieve satisfactory performance. We next applied the pipeline on newly acquired video to classify facial expression in real time. The testing showed that the pipeline can obtain good results over a range of imaging conditions.