An Adaptive Frame Selection and Deep Learning-Based Dynamic Hand Gesture Recognition System for Sterile Environments

Medical professionals such as doctors, surgeons, and medical lab technicians working in sterile environments cannot use their fingers on touch screens or other interfaces. For such environments, touchless interfaces can be proposed by using dynamic hand gesture recognition, where the doctors can study the patient's images by making different hand gestures in front of a camera. This article proposes a smart hand gesture-controlled image gallery application for sterile environments such as operation rooms, medical laboratories, etc. A 3DCNN-based deep learning model has been thus developed and trained on four selective gestures from the IPN hand gesture dataset, which are compatible with our application. An adaptive frame selection methodology has been developed in order to address memory and time constraints. The system has reached an accuracy level of 86.03%, which shows that it is well suitable for such an application.