User:Ke CHEN/Proposed/Gait recognition
With the increasing demands of visual surveillance systems, human identification at a distance is an urgent need. Gait is an attractive biometric feature for human identification at a distance, and recently has gained much interest from computer vision researchers. Gait is a particular way or manner of moving on foot. Compared with those traditional biometric features, such as face, iris, palm print and finger print, gait has many unique advantages such as non-contact, non-invasive and perceivable at a distance.
Contents |
A Framework for Gait Recognition
A typical gait recognition system is shown in Figure 1. Video data is firstly captured by a camera, and then walking persons are detected and segmented from the background by some motion detection and segmentation methods. After silhouettes are segmented from the background, gait features can be extracted. Human identification can be achieved by measuring the similarity between the extracted gait feature and those in a gait database.
Normally gait features are not robust enough to view variant, clothing changing, carrying bags, etc. To extract robust and discriminative feature is an important step in gait recognition. Here we will focus on gait feature extraction.
Gait Features
Databases and Evaluation
Even though many gait recognition algorithms have been proposed, comparison of different algorithms and evaluation of an algorithm's robustness to some variations such as the variations of view angle, clothing, shoe types, surface types, carrying condition, illumination, and time are still hard and open problems. These variations should be fully studied to develop robust and accurate gait recognition algorithms.
The HumanID Gait Challenge Problem
The HumanID Gait Challenge Problem \cite{usf:db:pami}, which consists of a large database, a baseline algorithm and twelve experiments, tried to handle these problems. The data in the HumanID Gait Challenge Problem was collected in an outdoor environment with complex background, The twelve experiments were designed to evaluate an algorithm's robustness to view, shoe, surface, time, clothing and carrying condition changes.
CASIA Gait Database and Evaluation Metrics
In the CASIA Gait Database there are three datasets: Dataset A, Dataset B (multi-view dataset) and Dataset C (infrared dataset).
Dataset A (former NLPR Gait Database) was created on Dec. 10, 2001, including 20 persons. Each person has 12 image sequences, 4 sequences for each of the three directions, i.e. parallel, 45 degrees and 90 degrees to the image plane. The length of each sequence is not identical for the variation of the walker's speed, but it must ranges from 37 to 127. The size of Dataset A is about 2.2GB and the database includes 19139 images.
Dataset B is a large multi-view gait database, which is created in January 2005. There are 124 subjects, and the gait data was captured from 11 views. Three variations, namely view angle, clothing and carrying condition changes, are separately considered. Besides the video files, we still provide human silhouettes extracted from video files.
Dataset C was collected by an infrared (thermal) camera in Jul.-Aug. 2005. It contains 153 subjects and takes into account four walking conditions: normal walking, slow walking, fast walking, and normal walking with a bag. The videos were all captured at night.
Other Databases
| Database Name | Num. of Subjects | Num. of Sequences | Environment | Time | Variations |
|---|---|---|---|---|---|
| UCSD Database cite{ucsd:database} | 6 | 42 | Outdoor | 1998 | - |
| MIT AI Database cite{lee:ellipsoidal} | 24 | 194 | Indoor | 2001 | View, time |
| Georgia Tech Database cite{gatech:web} | 20 | 188 | Outdoor, indoor, magnetic tracker | 2001 | View, time, distance |
| CMU Mobo Database cite{cmu:web,cmu:db} | 25 | 600 | Indoor, treadmill | Mar. 2001 | 6 viewpoints, speed, carrying condition, incline surface |
| HID-UMD Database（Dataset 1） | 25 | 100 | Outdoor | Feb.-May 2001 | 4 viewpoints |
| HID-UMD Database（Dataset 2） | 55 | 220 | Outdoor | June-July 2001 | 2 viewpoints |
| Soton Small Database cite{soton:web} | 12 | - | Indoor, green chroma-key backdrop | Carrying condition, clothing, shoe, view | - |
| Soton Large Database cite{soton:db,soton:web} | 115 | 2,128 | Indoor, outdoor, treadmill | Summer, 2001 | View |
| Gait Challenge Database cite{usf:db:pami,usf:web} | 122 | 1,870 | Outdoor | May and Nov. 2001 | 2 viewpoints, surface, shoe, carrying condition, time |
| CASIA Database（Dataset A） cite{cbsr:web} | 20 | 240 | Outdoor | Dec. 2001 | 3 viewpoints |
| CASIA Database（Dataset B） cite{yu:casiadb,cbsr:web} | 124 | 13,640 | Indoor | Jan. 2005 | 11 viewpoints, clothing, carrying condition |
| CASIA Database（Dataset C） cite{cbsr:web} | 153 | 1,530 | Outdoor, at night, thermal camera | Jul.-Aug. 2005 | Speed, carrying condition |
| TUM-IITKGP Database | 35 | 840 | Indoor | 2010 | hand-in-pocket, backpack, gown, static occlusion, dynamic occlusion |
| TUM-GAID Database | 305 | 3,370 | Indoor, Kinect, Audio + Image + Depth | Jan. and Apr. 2012 | backpack, coating shoes, time |
