A Sequenced Edge Grid Image Technique for Sign Language Recognition

Abstract

Currently there are many techniques and methods continuously proposed by researchers for Sign language recognition systems based-on machine learning. For data preprocessing for sign language, majority of researchers use single image of hands like static gesture images. Using only static hand images may not be efficient for real-world applications. In this paper, we propose an innovative technique for video processing called Sequenced Edge Grid Images (SEGI) for Sign Language recognition. The proposed SEGI is composed of images that represent the movement of hands within a single image, which can be applied to recognize a word or a sentence. To proof the concept, we have done several experiments with Thai sign language data collected from internet. SEGI was with existing techniques. Data are the Thai sign language learning video clips that are vocabularies to use in daily life. The proposed technique was implemented with convolutional neural network (CNN). For normal CNN, the experiments show that the SEGI with CNN increases of test accuracy rate with approximately 27% when compared to static gesture images.