Abstract Views :294 |
PDF Views:3
Authors
Affiliations
1 Cairo University, Computer Engineering Department, Giza, EG
2 Higher Thebes Institute for Engineering, Thebes Academy, Cairo, EG
Source
Digital Image Processing, Vol 3, No 17 (2011), Pagination: 1100-1106
Abstract
This paper introduces a new prototype system of sign to voice recognition technique to recognize Arabic signs and converts them into voice correspondences to enable Arabian deaf people to interact with normal people. The proposed technique captures a color image for the hand gesture and converts it into YCbCr color space that provides an efficient and accurate way to extract skin regions from colored images under various illumination changes. Prewitt edge detector is used to extract the edges of the segmented hand gesture. For its fast training process, probabilistic neural network (PNN) is used at classification stage where it uses a supervised training set to develop distribution functions within the pattern layer. These functions, in the recall mode, are used to estimate the likelihood of an input feature vector being part of a learned class. The nearest class with maximum score is selected and the corresponding sound clip is played. The proposed technique is used to recognize Arabic sign language alphabets and the most common Arabic gestures. Specifically, we applied the technique to 106 different signs and gestures with an average accuracy of 97.5% for three different signers at different situations. The proposed technique was also applied successfully to recognize Arabic fingerspelling. The detailed of the proposed technique and the experimental results are discussed in this paper.
Keywords
Arabic Sign Language, Fingerspelling, Gesture Recognition, Probabilistic Neural Network, Sign-to-Voice, Skin Color Segmentation.
