Elsevier

Image and Vision Computing

Volumes 48–49, April–May 2016, Pages 1-13
Image and Vision Computing

Viewpoint-independent gait recognition through morphological descriptions of 3D human reconstructions*

https://doi.org/10.1016/j.imavis.2016.01.003Get rights and content

  • We propose a new model-free approach for gait recognition.

  • The recognition is achieved independently of the trajectory.

  • Our method is based on 3D morphological analysis of gait sequences.

  • Our approach is able to identify people walking on curved paths.

Abstract

Many studies have confirmed gait as a robust biometric feature for identification of individuals. However, direction changes cause difficulties for most of the gait recognition systems, due to appearance changes. This study presents an efficient multi-view gait recognition method that allows curved trajectories on unconstrained paths in indoor environments. The recognition is based on volumetric analysis of the human gait, to exploit most of the 3D information enclosed in it. Appearance-based gait descriptors are extracted from 3D gait volumes and temporal patterns of them are classified using a Support Vector Machine with a sliding temporal window for majority voting. The proposed approach is experimentally validated on the “AVA Multi-View Dataset (AVAMVG)” and on the “Kyushu University 4D Gait Database (KY4D)”. The results show that this new approach is able to identify people walking on curved paths.

Introduction

Research on human gait as a biometric feature for identification has received a lot of attention due to the apparent advantage that it can be applied discreetly on the observed individual without needing the active participation of the individual.

Previous studies on gait recognition have been classified into two categories: model-based and model-free approaches. The model-based methods extract gait features by fitting a model to input images, whereas model-free approaches characterize the human gait pattern by a compact representation, without having to develop any articulated model for feature extraction and having practical application even with low quality images where the color and texture information is lost.

In addition, regarding viewing angle, the previous work can be categorized into two approaches: view-dependent and view-independent approaches. The view-dependent approaches assume that the appearance will not change during walking. In such methods, a change in the appearance, caused by a view change, will adversely affect performance [1]. Fig. 1 shows the influence of a curved path on the silhouette appearance. As one of the advantages of gait as biometric is that it does not need the cooperation of the individual, the trajectory of motion cannot be restricted to just straight paths.

On the other hand, the use of volumetric information allows more information to be analyzed in contrast to methods which only compute gait descriptors from silhouettes or 2D images. This paper presents an efficient view-independent method to recognize people walking along unconstrained (curved and straight) trajectories. This approach focuses on capturing 3D morphological and structural information from volumetric reconstructions of walking humans, which are previously aligned along the way. The main contribution is that our method allows direction changes, achieving a good recognition rate on unconstrained paths.

Some potential applications of this work are access control in special or restricted areas (e.g. military bases, governmental facilities and laboratories) or smart video surveillance (e.g. bank offices) [2].

This article is organized as follows. Section 2 describes works related to the topic of gait recognition. Section 3 explains the details of the proposed algorithm and describes three new descriptors which obtain information from 3D occupancy volumes. An analysis of the proposed method and the performance is given in Section 4. Finally, we conclude this paper in Section 5.

Section snippets

Related work

The previous work can be categorized into two approaches: view-dependent and view-independent approaches. In the following we describe works related to both categories.

Proposed method

We propose a model-free approach to recognize walking humans independently of the viewpoint and regardless direction changes. Our approach focuses on capturing 3D morphological and structural information from the gait through volumetric reconstructions of the walking humans.

The use of volumetric information allows more information to be analyzed in contrast to other related works, which only compute gait descriptors from silhouettes, discarding an important part of the dynamical and structural

Experiments and discussion

In this section we start by describing the used datasets, and then we present the experimental results.

Conclusions

This paper has proposed a method to recognize walking humans independently of the viewpoint and regardless direction changes. The method focuses on capturing 3D morphological and structural information from volumetric reconstructions of the gait. The main contribution is that the method achieves a good recognition rate on completely unconstrained paths, allowing direction changes, in contrast to others view-independent approaches where the view change is restricted to a few angles. In our

Acknowledgments

This work has been developed with the support of the Research Projects called TIN2012-32952 and BROCA both financed by Science and Technology Ministry of Spain and FEDER.

References (44)

  • S. Yu et al.

    Modelling the effect of view angle variation on appearance-based gait recognition

  • W. Hu et al.

    A survey on visual surveillance of object motion and behaviors

    IEEE Trans. Syst. Man Cybern. Part C Appl. Rev.

    (2004)
  • A. Kale et al.

    Gait analysis for human identification

  • L. Wang et al.

    Silhouette analysis-based gait recognition for human identification

    IEEE Trans. Pattern Anal. Mach. Intell.

    (2003)
  • S. Hong et al.

    Automatic gait recognition using width vector mean

  • S.D. Mowbray et al.

    Automatic gait recognition via Fourier descriptors of deformable objects

  • S.D. Choudhury et al.

    Silhouette-based gait recognition using Procrustes shape analysis and elliptic Fourier descriptors

    Pattern Recogn.

    (2012)
  • J. Han et al.

    Individual recognition using gait energy image

    IEEE Trans. Pattern Anal. Mach. Intell.

    (2006)
  • C. Rougier et al.

    Depth energy image for gait symmetry quantification

  • M. Hofmann et al.

    2.5D gait biometrics using the depth gradient histogram energy image

  • S. Sivapalan et al.

    Gait energy volumes and frontal gait recognition using depth images

  • S. Sivapalan et al.

    The backfilled GEI—a cross-capture modality gait feature for frontal and side-view gait recognition

  • Cited by (13)

    View all citing articles on Scopus
    *

    This paper has been recommended for acceptance by Mark S. Nixon.

    1

    Computing and Numerical Analysis Department, Edificio Einstein, Campus de Rabanales, Córdoba University, 14071 Córdoba, Spain. Tel.: +34 957212255.

    View full text