Classifier for Sinhala fingerspelling sign language

Abstract

Computer vision based sign language translation is usually based on using thousands of images or video sequences for model training. This is not an issue in the case of widely used languages such as American Sign Language. However, in case of languages with low resources such as Sinhala Sign Language, it’s challenging to use similar methods for developing translators since there are no known data sets available for such studies. In this study a sign language translation method is developed using a small data set for static signs of Sinhala Fingerspelling Alphabet. The classification model is simpler in comparison to Neural Networks based models which are used in most other sign language translation systems. The methodology presented in this study decouples the classification step from hand pose estimation and uses postural synergies to reduce dimensionality of features. This enables the model to be successfully trained on a data set as small as 122 images. As evidenced by the experiments this method can achieve an average accuracy of over 87% . The size of the data set used is less than 12% of the size of data sets used in methods which have comparable accuracies.