Pyramidal Fisher Motion for Gait Recognition
Francisco M. Castro, Manuel
J. Marín-Jiménez, Rafael Medina-Carnicer
Overview
The goal of this paper is to identify individuals by analyzing their gait. Instead of using binary silhouettes as input data (as done in many previous works) we propose and evaluate the use of motion descriptors based on densely sampled short-term trajectories. We take advantage of state-ofthe- art people detectors to define custom spatial configurations of the descriptors around the target person. Thus, obtaining a pyramidal representation of the gait motion. The local motion features (described by the Divergence-Curl-Shear descriptor) extracted on the different spatial areas of the person are combined into a single high-level gait descriptor by using the Fisher Vector encoding. The proposed approach, coined Pyramidal Fisher Motion [1], is experimentally validated on the recent ‘AVA Multiview Gait’ dataset [2]. The results show that this new approach achieves promising results in the problem of gait recognition.
Who are they?
The goal of this work is to identify people by using their gait. We build the Pyramidal Fisher Motion descriptor from trajectories of points. We represent here the gait motion of four different subjects.
Our approach
 |
a) The input is a sequence of video frames. b) Densely sampled points are tracked. c) People detection helps to remove trajectories not related to gait. d) A spatial grid is defined on the person bounding-box, so features are spatially grouped to compute a descriptor per cell. Then, those descriptors are concatenated into a single descriptor.
Downloads
| Filename | Description | Size |
|---|---|---|
| Source code for PFM | Coming soon | --- |
| Data+code: tracks of people | Precomputed tracks of people for AVAMVG and code to select dense trajectories close to people. | 16 MB |
| AVAMVG dataset | Website of AVAMVG dataset | --- |
Related Publications
[1]
F.M. Castro, M. Marin-Jimenez, R. Medina-Carnicer
Pyramidal Fisher Motion for Gait Recognition
Intl. Conference on Pattern Recognition (ICPR), 2014
Document: PDF
[2]
D. Lopez-Fernandez, F.J. Madrid-Cuevas, A. Carmona Poyato, M. Marin-Jimenez and R. Muñoz-Salinas
The AVA Multi-View Dataset for Gait Recognition (AVAMVG)
Intl. Workshop on Activity Monitoring by Multiple Distributed Sensing (AMMDS), 2014
Document: PDF via DOI |
Dataset: Go
Acknowledgements
This work has been partially funded by the Research Projects TIN2012-32952 and BROCA, both financed by FEDER and the Spanish Ministry of Science and Technology.