The Research Of Gait-Based Human Identification

Gait is a relatively new and emergent behavioral biometric,which pertains to the use of an individual's walking style to determine or validate identity,and gait-based human identification is a complex and challenging problem.To operate successfully,most of the established biometrics usually require proximal sensing or physical contact.However,they are hardly applicable at distance.Fottunately,gait,the way people walk,is still visible and can be easily perceived unobtrusively.So,from a surveillance perspective,gait is a very attractive modality.In this dissertation ,we do research on the key issues of gait-based human identification in two conditions:(1) Put forward an algorithm of gait-based human identification on the base of fusion and hidden markov model.We use the width vector of lower limbs as the gait character,and do periodical analyse on the human's width signal. We can get the initial frame of the sequence and exemplar by account the mean of subsection of every period at the same time. Aim at the characteristic of gait sequence,we analyse the feasibility of the hidden markov model,and put forward the method that use the exemplar as hidden markov model's state and the distance between the exemplar and the gait character of every frame as the model's observed vector. Experiment shows that the algorithm can achieve good results .To improve the identify rate further,we fuse the result of the recognise using muti-dimension gait character. Experiment shows that the method can achieve better results than before.(2) Put forward a new algorithm of gait-based human identification apply to multi-view environment. Accordingly,gait-recognition algorithms work best when presented with images where the person walks parallel to the camera. However, it is not realistic to expect that this assumption will be valid in most real-life scenarios. Hence it is important to develop methods whereby the side-view can be generated from any other arbitrary view in a simple and accurate manner. We show that it is possible to synthesize a side view from any other arbitrary view by using the perspective projection model and optical flow based structure from motion equations.It is no need to calculate depth,and apply to the entironment with a single camera. Experiment shows that the result of identification using synthesized gait sequence is beter than that using original sequence .