Gesture Recognition for Indian Sign Language using Kinect

Indian Sign Language static gesture depth dataset

Download the dataset from the following link.

https://drive.google.com/drive/folders/0B6iDOaIw70SceUFWQ0NoREVIUTA?resourcekey=0-fjQdHEkRhuPpOnlbICy2bg&usp=sharing

Please cite the following paper if you use this dataset in your research.

ANSARI, Z.A. & HARIT, G. Sadhana (2016) 41: 161. https://doi.org/10.1007/s12046-015-0405-3

You can read it for free at http://www.ias.ac.in/article/fulltext/sadh/041/02/0161-0182

This dataset consists of depth and RGB images (taken by Kinect XBox 360) of 140 sign language gestures of the Indian Sign Language taken from 18 subjects. There are a total of 5041 images each of resolution 640x480 in the depth and the RGB subsets. Each depth image has a corresponding RGB image.

Related video: https://www.youtube.com/watch?v=2oqD-_UCHxQ

This work was a part of my Master's thesis at Indian Institute of Technology, Jodhpur (2013).

Files have the name format: 'USER-XY-ABC-VZ.png' or 'USER-XY-ABC-VZ.txt' XY, ABC, VZ are numbers. XY represents volunteer number. ABC represents symbol number. VZ represents trial number of the corresponding volunteer. These numbers may be of 1, 2 or 3 digits. Trial numbers may be discontinuous.

User Number of Valid images
1       863
2       237
3       251
4       252
5       261
6       246
7       241
8       256
9       240
10      239
11      217
12      258
13      231
14      219
15      230
16      259
17      261
18      280

Symbols' meanings are provided in the vocabulary folder according to their number.

Each depth image is a matrix of size 640x480 having values in the range [0,2047] which correspond to real world depths according to the Kramer formula [1].

[1] Kramer J, Parker M, Herrera D, Burrus N and Echtler F 2012 Hacking the kinect. Apress