Human Action Recognition System In Video Image Sequence

Recently, as a active research topic, the study of vision-based human actionrecognition has attracted increasingly attention in computer vision and pattern recognition,which is related to many special fields such as pattern recognition, image processing,computer vision and artificial intelligence and so on. The aim of human action recognitionis attempting to detect, identify and track human bodies from image sequences, and moregenerally, to understand and describe their behaviors. The critical technology among themis concentrated on feature extraction and recognition algorithm.The study of human action recognition is a valuable theoretical problem and a hightechnique, which produces good economic value and social beneficial result, at the sametime, which has a wide range of potential applications, such as intelligent visualsurveillance, perceptual interface, virtual reality, human-computer interface, humanmovement analysis, etc. This paper launches the research around the related problems ofvision-based human action recognition, mainly including moving object extraction, featureextraction and algorithm of human action recognition three aspects. During the operationof human action recognition, the detection and tracking of moving human in video imagesequences belong to the research foundation of vision-based human action recognition,and the emphasis is carried out the researches on efficient feature extraction method andclassification selection. This article firstly introduces the operation platform ofexperimental simulation and some technology and arithmetic about processing image usedin this experiment. Meanwhile briefly gives some introduction of image acquisition, imagepreprocessing, moving objects detection and moving targets tracking in video imagesequences. The next step gives details information about the algorithms of featureextraction addressed the core technologies applied in this topic, then, presents the analysison human action identification and classification. Meanwhile, shows the explicit methodwe employed, and the effective analysis on experiment results. The last part summarizesthis dissertation and outlook the further study in this field.The research of this article is based on single individual in single scenes of popular action recognition databases employed accumulated motion image, Hu invariant moments,and kinematic feature, and combined with support vector machine as a classifier todistinguish the current activity in order to reach up to high recognition rate.