Facial Expression Analysis And Identification System

With the development of artificial intelligence in the life of people more and more extensive influence, how to implement the effective human-computer interaction, so that the machine can read people’s meaning and effective response has become an urgent problem to be solved. As the artificial intelligence is a very important branch, facial expression recognition has become an important research direction of artificial intelligence. In the study of facial expression recognition process, we use a variety of algorithms. In most cases the recognition rate and speed become the measure of these algorithms superiority is an important standard to judge.This paper studies a new automatic facial expression recognition algorithm, we first performed on the input image pretreatment, then feature points automatic positioning, the final classification tree strategy for identification. Compared with the classical algorithm, in the same database can get better recognition effect and speed. The main work and innovation are as follows:1) Introduces domestic and foreign scientific research institutions for research and development of the facial expression recognition, facial expression recognition is divided into two main components, namely the facial expression feature extraction, facial expression of category judgments.2) Image pretreatment and rapid positioning eye area, determine the position of a pupil, as active shape models provide initial position.3) The active shape model (active shape model) is the main principle. During training on the feature selection is improved, the redundant information of feature points for screening, reduces the calculation amount, improve the recognition speed. In the search stage by using a search window instead of the normal direction of search, reduces the edge region of the interference, and improve the recognition rate, and the experimental results show that the improved method is effective. 4) According to the weights of feature point selection method, the face is divided into several important areas, according to the different regions to identify each kind of expression contribution is different, adopt classification tree strategy for the identification of facial expressions, first rough classification, according to the upper level of the classification results of feature selection, the next level of the fine classification in each level, in the process of recognition with improved K nearest neighbor algorithm,Using classification trees strategy, first by a judgment result to determine the next step distance calculation, can effectively reduce the amount of calculation, improve the recognition speed.