The Research And Implementation Of AU Recognition And Expression Association Rule Mining Based On Deep Learning

Expression is the important of human heart activity.Through facial expression, human cognition can judge each other's attitude and inner world, expression is an important link, interpersonal human facial expression still lurking in the evolutionary history of human, so to explore and understand the mystery of the expression is also becoming more and more important.In recent years, many scholars have study the expression, such as the field of psychology study psychology factor of expression and judgment skills, physiology study expression in the field of physiological factors, but in today's information technology increasingly developed, expression has become an important topic of the intelligent human-computer interaction, which explore under the background of big data in the view of the computer facial expression is also one of the forefront of research.Deep learning is currently the latest artificial intelligence algorithm,This article from how to reveal and digging facial expression law of China between different ethnic groups, based on deep learning algorithm of face and expression data to analyze and identify the facial expressions between clan system, and then explore the multi-ethnic between the commonality and difference of facial expression.In this paper, the research content is as follows:1. Build a 3D face expression between clan system coordinates database.To improve the overall recognition rate, this paper this article obtains from the construction of database.Compared with collection of two-dimensional image for facial expression recognition, facial expression 3D coordinate point collection cloud for training can effectively eliminate the background information, skin texture information, and many other factors unrelated to facial expressions, greatly improve the rate of subsequent identification work.2. Select the series of the change of the AU three dimensional vector.Human facial expression is composed of a number of facial action unit namely AU, facial expressions vary but expression of AU is fixed, so the study of human needs from AU to research the mysteries of facial expression.This part adopts the maximum correlation minimum redundancy algorithm to choose the series of three dimensional vector of the change of the AU.3. Facial expression recognition based on depth of learning.In two parts on the premise, on the basis of this part of the building can study characteristics, the unity of the feature selection and classifier model tuning, finally choose to study through iteration between the AU, basic expressions, in line with the expression.