GaHu-Video: Parametrization system for human gait recognition
Description
This video dataset comprises (44) recordings of the human gait with (396) edited videos, there is also available a dataset of geometric features with (1431) files acquired from the video dataset that can be used for identifying people by human gait or for analyzing gait on various purposes. Each directory has the following structure: One directory with original videos of 44 people walking. Some parameters are, Video format AVI, stable background, resolution of 1280x700 p. Each video recording registers a person walking from left to right and back three times. One directory (Track A) with 44 edited videos with the first walking pass from left to right and back. A subdirectory (Sx_Track 1_Right) with 44 edited videos with the first walking pass from left to right. A subdirectory (Sx_Track 1_Left) with 44 edited videos with the first walking pass from right to left. This directory comprises 132 videos in total. One directory (Track B) with 44 edited videos with the second walking pass from left to right and back. A subdirectory (Sx_Track 1_Right) with 44 edited videos with the second walking pass from left to right. A subdirectory (Sx_Track 1_Left) with 44 edited videos with the second walking pass from right to left. This directory comprises 132 videos in total. One directory (Track C) with 44 edited videos with the third walking pass from left to right and back. A subdirectory (Sx_Track 1_Right) with 44 edited videos with the third walking pass from left to right. A subdirectory (Sx_Track 1_Left) with 44 edited videos with the third walking pass from right to left. This directory comprises 132 videos in total. One directory V-Geometric features with .dat files acquired from the video dataset with geometric features from the rectangle drawn over the silhouettes during a time. The geometric features were acquired with image processing techniques and are comprised of width, height, and area over time [1,2] nnW subdirectory: 477 .dat files with registered information about the behavior of the rectangle width during the gait record. nnH subdirectory: 477 .dat files with registered information about the behavior of the rectangle height during the gait record. nnA subdirectory: 477 .dat files with registered information about the behavior of the rectangle area during the gait record. References: [1] Senigagliesi, L.; Ciattaglia, G.; De Santis, A.; Gambi, E. People Walking Classification Using Automotive Radar. Electronics 2020, doi:10.3390/electronics9040588 [2] Kececi, Aybuke, Armağan Yildirak, Kaan Ozyazici, Gulsen Ayluctarhan, Onur Agbulut, and Ibrahim Zincir. 2020. Implementation of Machine Learning Algorithms for Gait Recognition. doi:10.1016/j.jestch.2020.01.005 [3] Figueiredo, Joana, Cristina P. Santos, and Juan C. Moreno. 2018. Automatic Recognition of Gait Patterns in Human Motor Disorders using Machine Learning: A Review. Vol. 53. doi: 10.1016/j.medengphy.2017.12.006
Files
Steps to reproduce
The video recordings are stored in AVI (Audio Video Interleave) format. The main purpose of having low-resolution videos is for using them as a part of embedded machine learning apps on hardware such as Raspberry Pi [3] or other IoT computational components. The video and data provided in this dataset can be used to extract new features or to train a machine learning system in order to identify people by human gait behavior or to analyze normal human gait parameters. A detailed description of this video dataset has been submitted to the journal Data in Brief, with the title "A video dataset of human gait with a dataset of geometric features for recognition purposes (GaHu-Video)".