Alphabet Recognition using Air written Trajectories

The enormous potential use of air-writing recognition in intelligent systems has made it highly popular. Some of the most fundamental issues in isolated writing are yet to be fully addressed. Writing a linguistic character or word in free space using a finger, marker, or handheld device is referred to as a trajectory-based writing method. It can be used where traditional pen-up and pen-down writing techniques are inconvenient. It has a significant upper hand over the gesture-based approach due to its simple writing style. However, because of the diverse characters and writing styles, it is a difficult process. In this paper, an alphabet recognition system for alphabets written in air, where the alphabet is recognised based on air trajectories which are three-dimensional (3D) and gathered by a single camera in this study. A reliable and effective colour-based segmentation is proposed to extract air recorded trajectories gathered by a standard web camera,. This solves the problem of push-to-write by removing limits on users' writing without the usage of an illusory box. The trajectory is normalized for improved recognition using convolutional neural network (CNN). We achieve recognition in real time with a high accuracy of 95% and negligible neural network complexity. It beats and surpasses the currently used techniques that involvewritten images as input.