Extraction of Hand Gesture Features for Indian Sign languages using Combined DWT-DCT and Local Binary Pattern

Abstract

The Hand Gesture system is based on two modes, viz, Enrollment mode and Recognition mode. In the enrollment mode, the Hand features are acquired from the camera and stored in a database along with the Sign languages. In the recognition mode, the hand features are re-acquired from the camera and compared against the stored Indian sign language data to determine the exact signs. In the pre-processing stage, two segmentation processes are proposed to extract the region of interest (ROI) of hand gesture. The first skin-color segmentation is used to extract the hand image from the background. The second region of interest of the hand gesture is segmented by using the valley detection algorithm. The Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT) are applied for the purpose of extracting the features. Further, the Sobel Operator and Local Binary Pattern (LBP) are used for increasing the number of features. The mean and standard deviation of DWT, DCT and LBP are computed.