Frontal-view Human Gait Recognition Based On Deterministic Learning And Kinect Stream

As a newly merged biometrics identification technology,gait recognition has become an attractive research field for computer vision and intelligent recognition,which aims to recognize human identification or detect physiological,pathological and mental characteristics by collecting,extracting,modeling,analyzing and classing based on human walking style.Compared with other static biometrics identification technology,such as fingerprint,iris and face recognition,gait feature is human explicit and dynamic representation,which is closely related to the information of spatial-temporal walking movement.Therefore,researching human gait recognition contains important meaning for its unique advantages and wide prospects.Gait recognition technology involved gait feature extraction,gait modeling and pattern recognition.Recently,gait recognition field has lots of researches and achieved so many excellent results,where most research works have collected human gait by video and side-view,and only concerned single kind of gait feature.But,in actual scene,frontal-view gait data is more convenient to be captured than side-view.Human gait is a very complex nonlinear kinetics signal,which contains rich and valuable feature information,only utilizing single-feature cannot fully represent human gait's attribute.It's very important to correctly model human gait and collect gait feature signal from frontal-view gait.Finally,exploring rich gait characteristic to promote gait recognition technology.This dissertation proposes a new frontal-view gait recognition based on deterministic learning(DL)and Kinect stream,DL can be apply to local accurate identification for uncertain dynamical gait and rapidly recognize dynamic gait.The main contribution and innovation of this dissertation are summarized as follows:1.Research human gait recognition in frontal-view,Kinect is applied to collect original human frontal-view gait data,and construct a original gait data base for later research work.Extract lots of gait feature and combined three single-features as a group,then experiment to obtain each group data's accuracy rate,and finally analyze human every single-feature's performance.2.Combined gait recognition technology with information fuse theory.To explore rich human gait feature and fuse very valuable gait information together to optimize input signal for gait system.This dissertation mainly research feature-level fusion and decision-level fusion,which can fully represent gait and greatly improve human gait recognition accuracy rate.Finally,verifying fusion algorithm theory by experiments and statistics analysis.3.Design and realize gait recognition system based on MATLAB software environment,which provide a friendly system interface and platform for researching gait and analyzing experiment.