Facial Expression Recognition Based On Random Forests

Facial expressions plays a very important role in people’s communication, it is an important of non-verbal communication way, because facial expression recognition have natural, intuitive, non-contact, safe, fast characteristics, so it cause wide public concern, it has become one of the most potential biometric identification technology, Currently, it is a hot research topic in the field of pattern recognition and data mining. This paper aims at to facial expression image as the research object, feature point positioning on face image, extracting the geometrical characteristics of the expression, to realize efficiently facial expression recognition, so as to build up a complete set of facial expression recognition model.This paper choose Japan JAFFE facial expression database, first preprocessing face image, including scale normalization and gray scale normalization, By using active shape model to achieve the automatic calibration of facial feature points of expression image, The corresponding characteristics of expression are accurately extracted, calibration mainly select jaw, mouth, nose, eyes, eyebrows, five major areas feature points of face, a total of 76, this paper analyze the feature points attribute, respectively statistics their geometric features, building the facial expression attribute table, the attribute table contains 42 geometric feature attributes, it provides the data support for subsequent facial expression recognition, With the method of random forests, through training face feature attribute data, Judgment on the importance of the characteristics, recognition on facial expression, and analyzing the classification results, so facial expression can be recognized more accurately and efficiently.