A hybrid approach to determine patients critical situation using expression & posture with convolutional neural network & Blazepose algorithm

Abstract

Patients are considered to be one of the most vulnerable persons. When it comes to critical patients their movements and behaviors need to be monitored constantly as simple negligences could result in severe consequences. It is almost impossible to monitor a patient 24/7 without making any slight error. Therefore, this paper will establish a simple but effective solution to this issue by creating a heuristic approach system that can detect a patient's facial expressions and postural movement to calculate the immediate conditions of patients with the assistance of deep learning algorithms. This is a hybrid approach as we have combined Convolutional Neural Network & BlazePose GHUM 3D to create a robust model which in our system can be used for image analysis in order to get precise monitoring results for critical situations by following specific sequences that would not have been possible without the hybrid model.