Font Size: a A A

Research On The Method Of Gait Feature Extraction And Recognition

Posted on:2007-07-27Degree:DoctorType:Dissertation
Country:ChinaCandidate:G J TianFull Text:PDF
GTID:1118360212467728Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Gait recognition, as a new biometric recognition method, has recently received growing interest within the computer vision community. HID (Human Identification at a Distance) is one of the major projects in DARPA (Defense Advanced Research Projects Agency) in the year 2000, combining the biometric recognition research findings of 26 American universities and companies, such as University of Maryland, MIT and CMU. The other biometric features e.g. (face, fingerprints) usually need short-distance apperception (fingerprints need contact scanner), so in the long-distance conditions, these methods are useless. Therefore, gait recognition would be the unique feature that can be perceived. On the basis of its importance both in theoretical research and practical application, gait recognition needs to be further studied. With the help of current research findings, this paper makes the following creative contributions:A Fourier Descriptors based gait contour feature representation method is proposed. The dynamic time warping technology is used to calculate the similarity between the gait sequences. This method of gait recognition is tested on CMU database and Little and Boyd database provided by DARPA. The correct recognition rate on Little and Boyd database is 3 % higher than that of the related reference but 10% lower on CMU database.The human body static shape features and dynamic features are combined based on Hidden Markov Model to recognize different people from their gaits. The correct recognition rate tested on the database provided by DARPA is 5% higher than that of the related reference.A gait recognition method based on human silhouette symmetry is proposed. The correct recognition rate tested on the database provided by DARPA is 5% higher than that of the related reference on small dataset; on large database it achieves 10%-25% higher to 100% correct recognition rate.A gait recognition method based on combined gait features is proposed. The correct recognition rate tested on the database provided by DARPA using the method...
Keywords/Search Tags:Gait Recognition, Fourier Descriptors, Dynamic Time Warping, Hidden Markov Model, Forward-backward Algorithm, Multi-covariates Time data, Feature Fusion, Feature Extraction, Feature match
PDF Full Text Request
Related items