Hindko Sign Language (HSL) Recognition Using Convolutional Neural Network

Abstract

Language has always played a significant role in Human to Human communication. In case of not knowing someone else's language, one can use hand gestures for communicating crudely but still be able to convey the message. Other than not knowing someone else's language there are millions of people in the world who have hearing or speaking disability. According to the World Health Organization (WHO), it has been estimated to be a population of 436 million (5% of the total world's population) people in the world who have hearing disabilities. Deaf people cannot use oral languages for communicating with other people. The source of their communication is Sign Language (SL) that conveys the message to the other person. In Computer Vision, there are different algorithms, which are used to interpret gestures and recognize them. The Deaf community of Pakistan uses its SL, like any other country in the world i.e., Pakistan Sign Language (PSL). There are around 60 local languages that are spoken in Pakistan including Hindko. Hindko and many other local languages spoken by minorities in Pakistan are on the brink of being endangered as the amount of research done on these languages is almost negligible. In this paper Convolutional Neural Network (CNN) is used for the recognition of Hindko Sign Language (HSL). Furthermore, we examine and analyze the recognition based on prediction to evaluate the efficiency of the utilized CNN. The methodology developed in this research work achieved an accuracy rate of 99.98%.