Facial Expression Analysis And Application With Local Characteristics And Machine Learning

Facial expressions are the foundation of communication between different people. It plays a very important role in everyone’s daily life. Facial expression recognition has caused attention widely over years. However, as the result of facial expression recognition is influenced by many issues such as facial features, emotional disclosure degrees and so on, facial expression recognition has long been a difficult task.The procedure of facial expression recognition can be mainly divided into two steps:feature extraction and classification. Feature extraction is to get the feature of expression out of facial images. There are two common ways to extract expression features. One of them is based on geometric relationship such as Facial Coding System while the other is based on appearance of performance such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and so on. In this research we use Local Binary Patterns (LBP) and Supervised Local Preserving Projection (SLPP) combined method for feature extraction. This method gets full access to texture information of the image, and then uses the information for dimension reduction. Experiments validate that feature extraction with this method has a higher recognition rate than the traditional methods.Besides, random forest is used in this research for classification. During classification, random forest selects the features randomly, which makes the demand for computing power reduced greatly. Experiments also validate that classification with random forest shows a good result.At last, facial expression recognition is used in the network teaching and facial expression analysis system. Recently, remote network teaching has been more and more popular. It usually starts with the students sending the request for learning and choosing the subject. Then the server plays the video of the subject. It is hard for the students to know their learning effect without teacher observing. In this situation, we propose the network teaching facial expression analysis system. The system allows students to track the status of their own learning at any time and find the knowledge that they haven’t understood. After that, they can learn again and again until they understand that. In this way, the learning efficiency is improved drastically. Besides, it also allows the system organizer to check the learning status of every subject and understand the difficulties about it.