Indian Sign Language gesture classification as single or double handed gestures

The development of a sign language recognition system can have a great impact on the daily lives of humans with hearing disabilities. Recognizing gestures from the Indian Sign Language (ISL) with a camera can be difficult due to complexity of various gestures. The motivation behind the paper is to develop an approach to successfully classify gestures in the ISL under ambiguous conditions from static images. A novel approach involving the decomposition of gestures into single handed or double handed gesture has been presented in this paper. Classifying gesture into these subcategories simplifies the process of gesture recognition in the ISL due to presence of lesser number of gestures in each subcategory. Various approaches making use of Histogram of Gradients (HOG) features and geometric descriptors using KNN and SVM classifiers were tried on a dataset consisting of images of all 26 English alphabets present in the ISL under variable background. HOG features when classified with Support Vector Machine were found to be the most efficient approach resulting in an accuracy of 94.23%.