Extensive feature set approach in facial expression recognition in static images

The article presents the preliminary concept of facial emotion recognition system. The approach focuses on the feature extraction and selection process which simplifies the stage of defining and adding new elements to the feature set. The evaluation of the system was performed with two discriminant analysis classifiers, decision tree classifier and four variants of k-nearest neighbors classifier. The system recognizes seven emotions. The verification step utilizes two databases of face images representing laboratory and natural conditions. Personal and interpersonal emotion recognitin was evaluated. The best quality of classification for personal emotion recognition was achieved by 1NN classifier, the recognition rate was 99.9% for the laboratory conditions and 97.3 for natural conditions. For interpersonal emotion recognition the rate was 82.5%