Gait Recognition Research Based On Multi-features Fusion Of Image Silhouette

Along with the rapid development of modern society, people have a growing emphasis on security issues. As an authentication technology, a biometric recognition which is for extremely strict safety requirements of the occasion required has attracted much attention by the recearcher. Traditional biometrics such as face, fingerprint, iris, usually require close or access to sexual awareness to complete the identification,which will be out of work in the case of long-distance. But gait recognition technology, as a new gait recognition biometric technology, has the characteristics of non-contact, easy collection that other traditional characteristics do not have, which overcome the traditional biometric require close contact and high-resolution images and other defects.Gait recognition has good recognition effect in the case of low resolution, which makes the gait recognition has broad application prospects in human behavior analysis and intelligent monitoring and so on.By reading the related literature of gait recognition technology at home and abroad, at present, duing to the model-based gait recognition algorithm is more complex and has a large amount of calculation; the researchers are used to adopting gait recognition method which are based on image contours. In the process of feature extraction,the contour is incomplete or only consider key frames, which caused the losing of part information and resulting in low recognition rate.In the process of selecting feature only consider single type characteristic, recognition rate is often not high.Therefore, in this paper, through the extraction of contour of two complete cycles of low dimension static characteristics-angle distance feature and dynamic feature-moment invariant features and frame difference percentage, which make the final features can better reflect the gait differences between individuals;at the same time,dynamic characteristics also avoid the impact of incomplete outline.During the experiment, this paper analysis the recognition effect of a single feature at fist, than analysis fusion feature again.By taking the nearest neighbor classifier,this paper achieves the gait classification decision.On this basis.In order to further improve the recognition rate, this paper has put forward the improved K neighbor method to classify recognition.First, by analyzing the insufficient of the common K neighbor method of the similarity of weight average, considering the weight problem, using the weights of different weighted mechanism.Then, constructs proper weight functionand analysis the process of selection of the weight function, determing the final weights of size.Finally, the improved K neighbor method is used to classify the gait feature.The experiment results compared with other algorithms, have confirmed the effectiveness of the proposed method.