Framework for Sinhala Sign Language Recognition and Translation Using a Wearable Armband

Sign Language is the main communication method of hearing & speaking impaired people and their main obstacle in the general society is the communication difficulty with normal hearing people. The aim of this research is to bridge above gap by proposing a framework for recognize Sinhala sign language gestures and translate them in to natural language. In order to preserve both functional and usability aspects of the solution, this study has used a non-invasive wearable gesture recognition armband. The approach is to use a combination of gestural data (surface Electromyography) that measures the muscle activity and spatial data (accelerometer, gyroscope & orientation) that measures hand movements for the sign recognition. The mapping has been done by implementing multiple artificial neural networks under the supervised machine learning technique. As a result, the study provided 100% accuracy for person dependent (personalized) study and 94.4% accuracy for person independent (generalized) study.