Automatic facial expression recognition system

Abstract

The paper is mainly about the idea of robots replacing humans in all aspects. Able to express is the only aspect that makes humans a step higher than the robots we produce. We project an idea where robot can mimic human expressions by means of the “automatic facial expression system”. Automatic facial expression recognition system that utilizes multi-stream Hidden Markov Models (NEURAL NETWORKS). The proposed system uses Facial Animation Parameters (FAPs), supported by the MPEG-4 standard, as features describing facial expressions. In particular, the FAPs controlling the movement of the outer-lips and eyebrows are used as visual features for classification. Experiments were performed under several different scenarios utilizing outer-lip and eyebrow FAPs individually and jointly. A new approach is proposed for introducing facial expression and FAP group dependent stream weights. The weights were chosen based on the facial expression recognition results obtained when FAP group streams are utilized individually. The proposed multi stream NEURAL NETWORKS facial expression recognition system achieves relative reduction of the expression recognition error of 44%, compared to the single-stream NEURAL NETWORKS system.