Abstract | ||
---|---|---|
A key-pose based gait recognition approach is proposed that utilizes the depth streams from Kinect. Narrow corridor-like places, such as the entry/exit points of a security zone, are best suited for its application. Alignment of frontal silhouette sequences is done using coordinate system transformation, followed by a three dimensional voxel volume construction, from which an equivalent fronto-parallel silhouette is generated. A set of fronto-parallel view silhouettes is, henceforth, utilized in deriving a number of key poses. Next, correspondences between the frames of an input sequence and the set of derived key poses are determined using a sequence alignment algorithm. Finally, a gait feature is constructed from each key pose taking into account only those pixels that undergo significant position variation with respect to the silhouette center. Extensive evaluation on a test dataset demonstrates the potential applicability of the proposed method in real-life scenarios. |
Year | DOI | Venue |
---|---|---|
2014 | 10.1007/978-3-319-16178-5_24 | COMPUTER VISION - ECCV 2014 WORKSHOPS, PT I |
Keywords | Field | DocType |
Gait recognition, Depth camera, Key pose, Incomplete cycle sequences, Variance image | Voxel,Coordinate system,Computer vision,Pattern recognition,Gait,Silhouette,Computer science,Artificial intelligence,Pixel | Conference |
Volume | ISSN | Citations |
8925 | 0302-9743 | 2 |
PageRank | References | Authors |
0.40 | 13 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Pratik Chattopadhyay | 1 | 52 | 6.81 |
Shamik Sural | 2 | 1008 | 96.36 |
Jayanta Mukhopadhyay | 3 | 72 | 26.05 |