Air Writing: Tracking and Tracing

Writing is an integral part of learning. Its value is quintessential in the development of cognitive understanding which aids in absorbing and de-mystifying the world around us. Air writing as a concept has been prevalent in education systems all across the world; teachers and parents have been training children to gesture the shapes of letters and spell out words in the air to build the habit of writing, but its technology is relatively recent. In this paper, we have introduced a method to implement air writing which allows users to trace linguistic characters in a pen-up pen-down motion of finger or hand thereby eradicating the need to write continuously. The proposed method is not frustrated by the requirement of extra hardware like a data glove or a specialized camera. The implementation makes use of the in-built camera in personal computers, detects the hand region, and traces or tracks the hand trajectory as per the user’s requirement. The implementation is broadly divided into two parts: (i) Detection of hand region; and (ii) Tracking and tracing hand movements. We propose the use of Single Shot Multibox Detector (SSD) with MobilenetV1 as the base network to detect the hand region. For the second half of the implementation, we have used the deque data structure and OpenCV. We have also included the option of complete/partial erasing of the screen to simplify the process of air writing and make it more intuitive.