INDIVIDUAL RECOGNITION BASED ON GAIT PATTERN

• Main Author: DESMOND BELAJA, DADU
• Format: Final Year Project Report
• Language: English; English
• Published: Universiti Malaysia Sarawak, (UNIMAS) 2022
• Subjects: T Technology (General)
• Online Access: http://ir.unimas.my/id/eprint/40245/1/Desmond%20Belaja%20Anak%20Dadu%2024pgs.pdf; http://ir.unimas.my/id/eprint/40245/4/Desmond%20Belaja%20Anak%20Dadu.pdf; http://ir.unimas.my/id/eprint/40245/
• Institution: Universiti Malaysia Sarawak

In this project, the focus is on human biometric system. Human biometric system can be divided into two groups which is physiological and behavioral. In this project, behavioral will be used, gait to be precise. Gait by definition is manner of walk, while Gait recognition is how to identify a human by using their walking style due to each person have different walking style. Convolution Neural Network, CNNs is the proposed method used for this project to classify the image of the gait via CNN Pre-Trained Network Model and to identify which pre-trained is the best for gait recognition. Gait Energy Image of the CASIA-B dataset in lateral view or equivalent to 90˚ of walk will be used. There are 124 subjects involve for CASIA-B dataset. Pre-Trained Network will be used to classify them. The main function of this Pre-trained Network is to classify each image and for feature extraction. There are a lot of pre-trained network that available out there, different pre-trained network will lead to different accuracy. As they have their own accuracy and speed for the image classification. In this project, MobileNets and ResNet-50 model will be used for the classification. Furthermore, Jupyter Notebook Software will be used to perform the image classification and feature extraction.