Gait Recognition Using Deep Convolutional Features


Pa, Pa Min and Sayeed, Shohel and Ong, Thian Song (2019) Gait Recognition Using Deep Convolutional Features. In: 2019 7th International Conference on Information and Communication Technology (ICoICT), 24-26 July 2019, Kuala Lumpur, Malaysia.

[img] Text
32.pdf - Published Version
Restricted to Repository staff only

Download (413kB)


Human gait recognition is the tracking of the people through their walking and recognizing human motion. Due to the benefit of remote access and un-cooperative subjects, gait recognition has gained attention in biometric-based identification. Gait recognition can be achieved by extraction distinct features by using hard-craft feature extraction and deep-learned features. Due to the disadvantage of hand-craft feature extraction that is not easily scalable for the diverse dataset, we proposed the convolutional neural network for feature learning in this paper. We presented simple ten layers’ convolutional network architecture with the Gait Energy Image as the input. Since gait recognition requires long training time, we use different activation functions in order to evaluate the result in term of classification accuracy and computational time. As for the activation functions, we utilized Rectified Linear Unit (ReLU), Leaky Rectified Linear Unit (LeakyReLU) and Parametric Rectified Linear Unit (PReLU) within the same CNN architecture. The experimental results indicate that the proposed method achieved the excellent recognition rate of 98.8% with less training time using CASIA-B Gait dataset.

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: Human activity recognition, gait recognition, convolutional neural network, activation function, deep learning, speed
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800-8360 Electronics
Divisions: Faculty of Information Science and Technology (FIST)
Depositing User: Ms Rosnani Abd Wahab
Date Deposited: 24 Aug 2021 13:51
Last Modified: 10 Apr 2023 04:02


Downloads per month over past year

View ItemEdit (login required)