Research On Learning Expression Feature Extraction Based On Improved Adaboost Algorithm In E-learning

The research in this subject is funded by the National Natural Science Foundation.(No:60970052)"The analysis of education expression research which is based on emotion recognition on e-learning system". And it also supported by Beijing Education Committee technology project, which No. is KZ200810028016and named as "Emotion recognition model based on personalized digital key technology education", and Beijing Natural Science Foundation, which No. is4112014and named as "The study of personalized e-learning community education based on emotional psychology'E-Learning as a modern mode of education in the network environment has been growing concern. However, most of the E-learning system lacks emotional interaction between students and teachers. The emotion missing situation is a serious impact on the effect of E-learning teaching. The main purpose of this thesis is for learning scenarios. It can get the academic expression feature information from the images of the students and use for providing the basis for academic emotions analysis and emotional cognitive teaching. The basic technical line is as following:By the camera installed on the machine, we can capture the learners'facial image. And it will detect learners' face image through image processing techniques. Then we can detect the facial features by the images which have been detected. and extract the parameters of its expression features. Which include the following parameters:the face area, interpupillary distance, eye spacing, eyebrow length, curvature of mouth and so on.The main research work and results of this thesis are as follow:(1) According to the requirements of the E-learning system real-time aspects of the learners face detection, study and improve the combination of the face detection algorithm based on skin color and the improved Adaboost method. The Adaboost algorithm is introduced in the classification of risk factors. Ensure that the algorithm can always give more attention to the positive samples. The experiment shows that the improvement combination algorithm is more in line with the requirements of the E-learning system, and gets better detection results.(2) According to the requirements of the E-learning system for learner academic expression parameters, based on gray scale smaller than a certain threshold algorithm, the iterative threshold selection algorithm. the Harris corner detection algorithm and eyebrows template method for learners pupillary distance, we can carry out improvements and applied research for learners pupillary distance, the spacing of your eyes, mouth curvature and eyebrows angle extraction. And achieve the expression parameter extraction which is required, and obtained good experimental results. (3) For the above algorithm and theory, we use Visual studio2008which calls the OpenCV technology to developed face detection and facial feather extraction system for E-learning. The experiment shows that the academic expression characteristic parameters which are extracted by the system can meet the needs of academic emotional modeling and emotional teaching of E-learning.