Gait recognition on the basis of markerless motion tracking and DTW transform

Abstract

In this study, a framework for view-invariant gait recognition on the basis of markerless motion tracking and dynamic time warping (DTW) transform is presented. The system consists of a proposed markerless motion capture system as well as introduced classification method of mocap data. The markerless system estimates the three-dimensional locations of skeleton driven joints. Such skeleton-driven point clouds represent poses over time. The authors align point clouds in every pair of frames by calculating the minimal sum of squared distances between the corresponding joints. A point cloud distance measure with temporal context has been utilised in k-nearest neighbours algorithm to compare time instants of motion sequences. To enhance the generalisation of the recognition and to shorten the processing time, for every individual a single multidimensional time series among several multidimensional time series describing the individual's gait is established. The correct classification rate has been determined on the basis of a real dataset of human gait. It contains 230 gait cycles of 22 subjects. The tracking results on the basis of markerless motion capture are referenced to Vicon system, whereas the achieved accuracies of recognition are compared with the ones obtained by DTW that is based on rotational data.

Review of gait recognition approaches and their challenges on view changes

“…Differently, the methods proposed in [77–82] performed the gait recognition based on 3D data collected from the proposed markerless 3D motion tracking system, which was validated using the commercial motion capture system from Vicon Nexus [77]. In [77], the tensorial gait data was optimised and classified using multi‐linear PCA.…”

Combined Regularized Discriminant Analysis and Swarm Intelligence Techniques for Gait Recognition

“…The grey wolves start the attack, once the prey stops moving. To model the process of approaching the prey, the GWO linearly decrease all the values of a from 2 to 0 according to the equation [32] a = 2 − 2t T max (19) where T max is the total number of iterations of the algorithm. The change of a affects the coefficient vector A, which controls the behavior of search agents.…”

“…Gait recognition methods [3] can be categorized as model-free (appearance-based) [2,[4][5][6][7][8][9][10][11][12] and model-based [13][14][15][16][17][18][19][20][21][22]. Most gait recognition studies are based on model-free approaches that employ the whole motion pattern of the human body.…”

“…In the recognition process, various classification techniques are used; most often, they are classical methods such as k-nearest neighbors (kNN) [2,4,10,[14][15][16][17]19,23], multilayer perceptron (MLP) [21], support vector machine [9,16,24] linear discriminant analysis (LDA) [16], and radial basis function neural networks [18]. The main focus in these papers is on developing descriptors that better describe gait features, rather than introducing new classification methods with better recognition ability.…”

Motion Recognition Algorithm Based on Deep Edge-Aware Pyramid Pooling Network in Human–Computer Interaction

“…First, the PR value is calculated by predicting the contour processing of the pedestrian, and then the PR curve is drawn. In order to further verify the classification accuracy of the improved deep network for motion dataset proposed in this paper, this paper selects the intelligent algorithms with more engineering applications for comparison: SRM [5], BMP [28], DSS [8], MSR [29], DCL [30]. For a fair comparison, the results of all comparison algorithms are derived from the original experimental results and source code.…”