Abstract:
Gesture recognition and sign language recognition have remained a considerably important research area. However, Pakistani Sign Language (PSL) failed to gather interest within the academia with respect to the other regional variants that include American Sign Language (ASL) and British Sign Language (BSL), to name a few. The proposed system uses static images to extract local and global, region and boundary-based descriptors for acquiring gesture information, which is provided as input for supervised learning method known as Support Vector Machine (SVM). Some of the primary features incorporated within this system include Hu invariant moments and Fourier descriptors. Meanwhile, the data collection was carried out with the help of sixty individuals in order to obtain 1,800 feasible images, which were used for the machine learning within this system purposes (1,000 images for training and 800 images for testing purposes).
The purpose of this research is to formally introduce a practical machine learning-based PSL recognition system, which can lay the groundwork for future research pertaining to PSL. While, more importantly, allowing the development of a system framework that significantly reduces the communication barrier between the Pakistani society and its marginalized deaf community. The proposed system demonstrates an accuracy of 74% for thirty-three PSL gesture classes corresponding to the selection of Urdu language alphabets. This research can be considered as an initial step into a long-term research plan towards transforming and creating a real-time PSL recognition system that is able to assist deaf individuals within their day-to-day lives that deals with static as well as dynamic gestures.
