Depth Based 3D Indian Sign language Recognition using Adaptive kernels
A.S.C.S. Sastry1, A. Ram Kishore2, Ch. Bulli Raju3, P.V.V. Kishore4, D. Anil Kumar5, E. Kiran Kumar6, M. Teja Kiran Kumar7

1A.S.C.S. Sastry, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
2A. Ram Kishore, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
3Ch. Bulli Raju, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
4P.V.V. Kishore, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
5D. Anil Kumar, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
6E. Kiran Kumar, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
7M. Teja Kiran Kumar, Department of ECE, KLEF, Deemed to Be University, Green Fields, Vaddeswaram, Guntur (Andhra Pradesh), India.
Manuscript received on 07 April 2019 | Revised Manuscript received on 20 April 2019 | Manuscript published on 30 April 2019 | PP: 914-918 | Volume-8 Issue-6, April 2019 | Retrieval Number: F3774048619/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: This work discusses a new approach for Indian sign language recognition (ISLR) using Depth sensor. We propose a novel approach for recognizing sign language gestures from an RGB, depth video sequences by extracting the histogram of oriented gradient (HOG) features and their recognition using adaptive kernel (AK) matching. The kernel-based methods are remarkably effective for recognizing the 2D and 3D actions. This work explores the potential of the adaptive kernels in fusion of RGB and depth kernel scores using ISLR from depth sensor. Accordingly, the HOG features were encoded into adaptive kernels. The recognition is carried out based on the similarity between the query and database features. The performance of the our approach tested on our own 100 class, 5 subjects sign data named as BVCSL3D, captured using Microsoft Kinect v2 sensor and two other publicly available action datasets NTU RGB-D and UTKinect. Our method outperforms when compared to other previous methods on the above datasets.
Keyword: Histogram of Oriented Gradient (HOG), Adaptive Kernels, Kinect Sensor, 3D Sign Language Recognition.
Scope of the Article: Pattern Recognition