Calibrated and synchronized multi-view video and motion capture dataset for evaluation of gait recognition

Bogdan Kwolek,Adam Switonski,Agnieszka Michalczuk,Tomasz Krzeszowski,Konrad Wojciechowski,Henryk Josinski

doi:10.1007/s11042-019-07945-y

Abstract

We introduce synchronized and calibrated multi-view video and motion capture dataset for motion analysis and gait identification. The 3D gait dataset consists of 166 data sequences with 32 people. In 128 data sequences, each of 32 individuals was dressed in his/her clothes, in 24 data sequences, 6 of 32 performers changed clothes, and in 14 data sequences, 7 of the performers had a backpack on his/her back. In a single recording session, every performer walked from right to left, then from left to right, and afterwards on the diagonal from upper-right to bottom-left and from bottom-left to upper-right corner of a rectangular scene. We demonstrate that a baseline algorithm achieves promising results in a challenging scenario, in which gallery/training data were collected in walks perpendicular/facing to the cameras, whereas the probe/testing data were collected in diagonal walks. We compare performances of biometric gait recognition that were achieved on marker-less and marker-based 3D data. We present recognition performances, which were achieved by a convolutional neural network and classic classifiers operating on gait signatures obtained by multilinear principal component analysis. The availability of synchronized multi-view image sequences with 3D locations of body markers creates a number of possibilities for extraction of discriminative gait signatures. The gait data are available at http://bytom.pja.edu.pl/projekty/hm-gpjatk/.

Highlights

Gait is a complex function of body weight, limb lengths, skeletal and bone structures as well as muscular activity
Gait-based person identification has been performed using Naıve Bayes (NB), Support Vector Machine (SVM), Multilayer Perceptron (MLP) and k-Nearest Neighbor (1NN, 3NN and 5NN) classifiers, operating on features extracted by the Multilinear PCA
Sequential Minimal Optimization (SMO) is one of the common optimization algorithms for solving the Quadratic Programming (QP) problem that arises during the training of SVMs

Summary

Introduction

Gait is a complex function of body weight, limb lengths, skeletal and bone structures as well as muscular activity. From a biomechanics point of view, human walk consists of synchronized, integrated movements of body joints and hundreds of muscles. Barclay et al [8] showed that the identity of a friend and the person’s gender can be determined from the movement of light spots only They investigated both temporal and spatial factors in gender recognition on the basis of data from point light displays. Benoit et al [11] investigated the effect of skin movement on the kinematics of the knee joint during gait through comparing skin markers to pin in bone markers They found that kinematic analysis is burdened with errors resulting from movement of the soft tissue, which should be considered when interpreting kinematic data. Marker-based moCap systems are costly and they are not available in many clinical settings [57]

Methods

Results

Conclusion