1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Human Gait Recognition: An Averaged Silhouette Based Approach
Authors:	Negin Karimi Hosseini (P53690)
Supervisor:	Md. Jan Nordin, Assoc. Prof. Dr.
Keywords:	Human Gait Recognition An Averaged Silhouette Based Approach Human Gait An Averaged Silhouette Based Approach Gait Recognition An Averaged Silhouette Based Approach Walking--Data processing
Issue Date:	15-May-2013
Description:	Fingerprints, iris or face biometrics are common human recognition methods that need subject assistance or physical contact. These methods cannot be used for distance human recognition and without subject assistance in a real world. Gait can be used as a human biometric to recognize subject by the way they walk without subject assistance. Gait recognition systems can be applied in surveillance environments or human machine interaction purposes. Recently, gait recognition has become an interest of researchers. Although various gait recognition systems are suggested by researchers, they are still far from the human’s ability to recognize people by the way they walk. The factors that are needed to be improved in future studies are: recognition rate, speed, accuracy, computational cost. Since, several characteristics can be achieved from the gait it is significant to find out which features or parts of the human body are most effective in gait recognition. We implemented a gait recognition system based on the silhouette of the subject in each frame. This type of gait recognition system is called appearance-based or model free gait recognition system. First, we applied a blob detection algorithm to detect the human subject and track it. Then, we estimate the gait cycle in the gait sequences by counting white pixels in binarized frames. After gait period estimation, we compute the Averaged Silhouettes of each subject which is a signature to represent the walking characteristics of the subjects. Afterwards, we applied Principal Component Analysis as the feature extraction technique which is also applied for the dimensional reduction purpose. We trained our algorithm with the training set. Finally, we compute the Euclidean distance for the recognition step. We examined the algorithm with two experiments. In the first experiment we used the complete Averaged silhouettes and in the second experiment we applied the upper part of the averaged silhouettes. The suggested algorithm is implemented on the normal walking gait sequences of the TUM-IITKGP Gait Database. The results illustrated that upper part of averaged silhouettes achieved better results in compare with the complete averaged silhouettes. These results demonstrate that the upper part of the silhouettes are more suitable for the feature extraction stage in this method of gait recognition.,Master
Pages:	99
Call Number:	TK7882.P3 .H647 2013 3
Publisher:	UKM, Bangi
URI:	https://ptsldigital.ukm.my/jspui/handle/123456789/476181
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.