Gait recognition using multichannel convolution neural networks
- PDF / 1,488,815 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 108 Downloads / 270 Views
(0123456789().,-volV) (0123456789().,-volV)
EXTREME LEARNING MACHINE AND DEEP LEARNING NETWORKS
Gait recognition using multichannel convolution neural networks Xiuhui Wang1
•
Jiajia Zhang1 • Wei Qi Yan2
Received: 30 October 2018 / Accepted: 5 October 2019 Ó Springer-Verlag London Ltd., part of Springer Nature 2019
Abstract Human gait recognition has a wide range of applications in multiple fields, such as video surveillance, digital security, and forensics. In this paper, we investigate the challenging problem of cross-view gait recognition and propose a novel gait recognition scheme by utilizing the strong expression of convolution neural networks (CNN). First, instead of using gait energy images in traditional gait recognition, we will design a new gait feature representation, trituple gait silhouettes, constructed by using consecutive gait silhouette pictures. Second, we will construct a multichannel CNN network to tackle a set of sequential images in parallel. Each of the image datasets is treated as one input channel, and a different convolutional kernel is used. Finally, the proposed approach is evaluated extensively based on the CASIA gait dataset A/B for cross-view gait recognition, and further on the OU-ISIR large population gait dataset to verify its generalization capability with large-scale data. To the best of our knowledge, this is the first time that this gait recognition scheme is presented. All our experimental results show that the proposed method obtains better performance when compared to those existing methods. Keywords Gait recognition Multichannel CNN Biometrics recognition Video-based surveillance
1 Introduction As one of the most promising biometrics, gait embodies human walking habit and its dynamic characteristics. Compared to other biometrics, gait has many advantages, such as non-offensive extraction and low-resolution compatibility [1, 2]. Recently, human gait recognition has gained tremendous attention, and a slew of approaches [3–5] have been proposed to improve gait recognition. Nowadays, the focus of existing methods mainly is on different view angles of gait, namely, cross-view problem. However, the problem is still far away from being completely solved because gait appearance of one person may be dramatically altered when the view angles are changed. This work was completed when the Xiuhui Wang was a visiting scholar with the Auckland University of Technology, New Zealand. & Xiuhui Wang [email protected] Wei Qi Yan [email protected] 1
No. 258, Xueyuan Street, Hangzhou 310018, China
2
No. 2-14, Wakefiled Street, Auckland 1010, New Zealand
Fortunately, with successful applications of deep learning (DL) [6] in object detection, segmentation, and recognition from images and videos, some researchers [7, 8] have applied deep learning to human gait recognition. As an outstanding representative with optimized network structure by combining local perceptions, weights sharing, and spatial downsampling to make good use of the characteristics of samples, convolution neural
Data Loading...
