Human Gait Recognition Based On Skeleton Model Classification

The gait recognition is a emerging technology in the field of biometric characteristics recognition and computer vision.It mainly inspect the human gait to identify the person's identity by characteristics of the walking gait.In recent years,this gait recognition technology has gained extensive interest of the associated specialists,as both the rising awareness of national security and the demands for intelligent monitoring system at some important places.Gait recognition has became the most potential biometric characteristics identification technology with its superior recognition method.Since it is only in the stage of theoretical research,a series of inevitably problems are still in urgent need to be solved in Relevant technology realization.Such as the error of algorithm to extract the foreground moving target is large,the difficult to track the gait feature points,the undesirability for the algorithm of characteristics information.All above afflictions are discussed deeply in this article,and a gait recognition algorithm is presented to solve above problems,which is based on the human body skeleton model classification.The corresponding researches cover the following three aspects of issues mainly:Firstly,In the aspect of foreground target segmentation,some algorithms are presented for the complex background information,which are to extract the foreground target,standardize the image and skeletonized the image.Secondly,In the aspect of gait characteristic extraction,the Five Connecting Rod Model of Human Movement is established.Then the parametric gait characteristics of human movement are obtained,based on matching between the model and the skeletal image with Neural Networks.And the Fourier Descriptor of gait characteristic is proposed to translate the characteristics from time domain to frequency domain,considering the gait information unequal in the time domain.Thirdly,In the aspect of gait characteristic classification,a progressive algorithm,based on the Projection Pursuit theory and the Mean Shift Algorithm is raised to decrease the number of the comparison samples in the process of KNN classifier classification and improve the classification efficiency.In the end,a gait recognition system is constructed with the algorithm of gait recognition researched based on C++ programming platform.And the system performance are tested by the experiment,and shows an important theoretical meaning and practical value at the aspect of image pre-processing,gait characteristic extraction and gait characteristic extraction.