Journal of Computer Applications ›› 2024, Vol. 44 ›› Issue (7): 2250-2257.DOI: 10.11772/j.issn.1001-9081.2023070977

• Multimedia computing and computer simulation • Previous Articles     Next Articles

Gait recognition algorithm based on multi-layer refined feature fusion

Ruihua LIU(), Zihe HAO, Yangyang ZOU   

  1. School of Artificial Intelligence,Chongqing University of Technology,Chongqing 401135,China
  • Received:2023-07-19 Revised:2023-09-30 Accepted:2023-10-07 Online:2023-10-26 Published:2024-07-10
  • Contact: Ruihua LIU
  • About author:HAO Zihe, born in 1999, M. S. candidate. Her research interests include machine vision, gait recognition.
    ZOU Yangyang, born in 1981, Ph. D., lecturer. Her research interests include machine learning.
    First author contact:LIU Ruihua, born in 1975, Ph. D., associate professor. His research interests include medical image segmentation, gait recognition.
  • Supported by:
    Chongqing Natural Science Foundation(CSTB2023NSCQ-MSX0319);Chongqing University of Technology Graduate Innovation Project(gzlcx20223457)

基于多层级精细特征融合的步态识别算法

刘瑞华(), 郝子赫, 邹洋杨   

  1. 重庆理工大学 两江人工智能学院,重庆 401135
  • 通讯作者: 刘瑞华
  • 作者简介:郝子赫(1999—),女,吉林长春人,硕士研究生,主要研究方向:计算机视觉、步态识别;
    邹洋杨(1981—),女,四川成都人,讲师,博士,主要研究方向:机器学习。
    第一联系人:刘瑞华(1975—),男,湖南永州人,副教授,博士,CCF会员,主要研究方向:医学图像分割、步态识别;
  • 基金资助:
    重庆市自然科学基金资助项目(CSTB2023NSCQ-MSX0319);重庆理工大学研究生创新项目(gzlcx20223457)

Abstract:

Deep learning has further promoted the research of gait recognition algorithms, but there are still some problems, such as ignoring the detailed information extracted by shallow networks, and difficulty in fusing unlimited time-space information of gait videos. In order to effectively utilize shallow features and fuse time-space features, a cross-view gait recognition algorithm based on multi-layer refined feature fusion was proposed. The proposed algorithm was consisted of two parts: Edge Motion Capture Module (EMCM) was used to extract edge motion features containing temporal information, Multi-layer refined Feature Extraction Module (MFEM) was used to extract multi-layer fine features containing global and local information at different granularities. Firstly, EMCM and MFEM were used to extract multi-layer fine features and edge motion features. Then, the features extracted from the two modules were fused to obtain discriminative gait features. Finally, comparative experiments were conducted in multiple scenarios on the public datasets CASIA-B and OU-MVLP. The average recognition accuracy on CASIA-B can reach 89.9%, which is improved by 1.1 percentage points compared with GaitPart. The average recognition accuracy is improved by 3.0 percentage points over GaitSet in the 90-degree view of the OU-MVLP dataset. The proposed algorithm can effectively improve the accuracy of gait recognition in many situations.

Key words: biometric recognition, gait recognition, feature extraction, gait silhouette sequence, spatiotemporal feature fusion

摘要:

随着深度学习的引入,步态识别算法取得了很大的突破,但是仍存在忽略了浅层网络提取的细节信息,以及对不限时长的步态视频时空信息难以融合的问题。为了有效利用浅层特征和融合时空特征,提出一种基于多层级精细特征融合的跨视角步态识别算法。所提算法由两个部分组成:边缘运动捕捉模块(EMCM)用于提取包含时间信息的边缘运动特征;多层级特征提取模块(MFEM)用于提取包含不同粒度全局和局部信息的多层级精细特征。首先,使用EMCM和MFEM分别提取多层级精细特征和边缘运动特征;然后,将两个模块提取的特征融合得到具有鉴别性的步态特征;最后,在公开数据集CASIA-B上和OU-MVLP上进行多种情况下的对比实验。在CASIA-B上平均识别准确率可达89.9%,与GaitPart相比,所提算法的平均识别准确率提升了1.1个百分点;在OU-MVLP上90°视角下比GaitSet识别准确率提升了3.0个百分点。所提算法能够有效地提升多种情况下的步态识别的准确率。

关键词: 生物特征识别, 步态识别, 特征提取, 步态轮廓序列, 时空特征融合

CLC Number: