Systematic Review on on-Air Hand Doodle System for the Purpose of Authentication

Abstract

Air-drawing for authentication or gesture recognition is vastly studied such that new methods for the character drawn can be identified with better accuracy and be independent of any hardware components such as gloves to detect the fingertip movement or any wearable sensor or using any external pen. This article surveys how air-writing is done. Some of them include using fingertips as a pen and having the trained CNN model against it using the techniques DNST and Bi-LSTM for the regression and hand gesture identification and classification. This system had an overall accuracy of 88 percentage. Another system solves the air-writing issue by using deep learning architecture, by executing it on 3D space where the next evolution is numerical digits structured into multiple dimensions time-series extracted from a sensor called LMC. In another system, wi-write recognizes hand writing mainly to overcome blurred vision and neurological diseases in people, this will use COTS WiFi but is a device-free handwriting recognition system. In many other systems deep learning models are used and those parameters can be used to achieve higher recognition rate for example 98 percentage above.