R.B. MAPARI1*, G.U. KHARAT2
1Anuradha Engineering College, Chikhli, 443201, MS, India
2Sharadchandra Pawar College of Engineering, Otur, 412409, MS, India
* Corresponding Author : rajesh_mapari2001@yahoo.com
Received : 13-05-2016 Accepted : 01-06-2016 Published : 14-06-2016
Volume : 7 Issue : 2 Pages : 474 - 479
Int J Mach Intell 7.2 (2016):474-479
Keywords : ASL, MLP, GFFNN, SVM
Academic Editor : Dr. Chetna Dabas, Dr. S S Nagamuthu Krishnan, Dr. Rajashree Shettar, Dr. K.s. Jasmine, Dr Pravin Ambadas Kharat, Dr Devpriya Soni
Conflict of Interest : None declared
Acknowledgements/Funding : None declared
Author Contribution : None declared
The American Sign Language (ASL) alpha-numeric character classification/recognition without using any aid (embedded sensor, color glove) is really difficult task. This paper describes a novel method to classify static sign by obtaining feature set based on DCT (Discrete Cosine Transform) and Regional properties of hand image. Feature set of size 1860×74 is later trained and tested using different classifiers like MLP, GFFNN, SVM. We have collected dataset (alpha numeric character) from 60 people including students of age 20-22 years and few elders aged between 25-38 who have performed 31 signs resulting in total dataset of 1860 signs. Out of this 90% dataset is used for training and 10% considered for Cross validation. We have got maximum classification accuracy as 86.16 % on CV dataset using GFF Neural Network.