Gait classification through CNN-based ensemble learning
- PDF / 1,989,974 Bytes
- 17 Pages / 439.642 x 666.49 pts Page_size
- 50 Downloads / 227 Views
Gait classification through CNN-based ensemble learning Xiuhui Wang1
· Ke Yan1
Received: 9 April 2020 / Revised: 26 August 2020 / Accepted: 28 August 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Gait is a biological characteristic for video surveillance and many other applications, which can be used to identify individuals at a large distance. In this paper, a gait classification framework based on CNN Ensemble (GCF-CNN) is proposed, which includes three modules: 1) Feature extraction and preprocessing: use random sampling with replacement strategy to generate a serial of training sets from gait silhouette images; 2) Gait models training: construct and train primary CNN classifiers using different hyper-parameters, and train them a secondary classifier to combine them; 3) Gait classification: utilize the trained two-level classifier to achieve gait classification. In addition, the proposed classification framework is evaluated on the CASIA Gait Database and OU-ISIR Gait Database. And it is demonstrated by comprehensive experiments that the proposed classification framework can achieve outstanding performance in correct classification rate with respect to several state-of-the-art methods. Keywords Ensemble learning · CNN · Gait recognition
1 Introduction Gait is an appealing biometric feature which can be used for human recognition at a distance. Compared with other means of biometric authentication like fingerprints or faces, gait recognition can be applied without alerting or disturbing the target subjects [3, 16, 31]. On the other hand, the research on efficient and practical gait classification methods still remains a formidable challenge and an area of active research, namely, most of the existing gait recognition algorithms only work well under the best condition of image and video acquisition [3, 31]. Recently, with the successful applications of Deep Learning (DL) technologies [9] in object detection, segmentation and recognition from images and videos [22, 23], some Xiuhui Wang
[email protected] 1
Key Laboratory of Electromagnetic Wave Information Technology and Metrology of Zhejiang Province, College of Information Engineering, China Jiliang University, No. 258, Xueyuan Street, Hangzhou, 310018, China
Multimedia Tools and Applications
researchers have preliminarily applied deep learning to human gait recognition [19, 31]. Convolution Neural Networks (CNN), as an outstanding representative of DL technologies, has many advantages such as combining local perceptions, weights sharing and spatial down-sampling to make full use of limited sample data. Besides, Ensemble Learning is one of the DL technologies which combines multiple primary learners through a fusion strategy to improve the overall generalization performance [28].Ensemble learning has attracted wide attentions due to its easily understandable structure and promising classification performance by combining primary learners into a stronger one. Elghazel et al. [2] proposed an ensemble method, Ran
Data Loading...
