Key Technology Research In Gait Recognition Based On Human Gait Contour

Human gait recognition is a biometric recognition that identifies individuals by their walking manners. Compare to other biometric recognitions, its advantages are that it is a long distance recognition technology, noninvasive and difficult to conceal. Therefore more and more researchers from different countries have high interest to this kind of recognition. Generally, gait recognition consists of three parts: processing gait sequence, feature extraction and feature classification. Under project supported by National Nature Science Foundation of China, the paper explores the key technology in gait recognition based on body contour in these three aspects:Firstly, the paper briefly introduces gait recognition technology and reviews predecessor's research on methods and approaches in gait recognition, then this paper summarizes the gait research status and progress. Moreover, the main work of this paper is also described.Secondly, three key technologies in gait recognition based on human gait contour are systemically studied and explored in this paper. In motion imagine sequence pretreatment section, the paper introduces the background subtraction technology. Background subtraction technology is the most common method in movement segmentation; it can provide reliable feature data. This paper adopts 2-D imagine following algorithm to get the contour of human motion imagine after the whole human motion area is obtained. In feature extraction section, we pick up four key motion movement frames after getting the contour of human motion imagine sequence. Shape feature is one of the most important low-grade feature in imagine, and it is also the easiest feature to get and the most intuitional feature. The shape express method based on gait contour edges can extract shape features by using edge information of subject directly. The paper explores the method of using Fourier descriptor to describe gait feature on the base of edge shape features. Fourier descriptor has three advantages: it can transform imagine into 1-D feature and it is not sensitive to start, rotation and shift; last, the most important advantage is that Fourier frequency component mostly locates on its low frequency part, so it reduce the dimension of recognition data greatly. In classification, the paper discusses the advantage and disadvantages of the classification methods using presently and studies the Nearest Neighbor Method from two main factor: parameter choice and fusion strategy. The paper utilizes the Nearest Neighbor Method in Euclidean Space to classify the gait features and the test has proved that this algorithm has correct classification result and high efficiency.Lastly, the work of this paper is summarized, the difficulties in the present research are examined and further research is pointed out..