Efficient Composite Model to Detect Facial Paralysis using Generative Adversarial Network and Facial Key Points Analysis

Abstract

Abstract Facial paralysis is the inability of one or both sides of the face's muscles to move, and it can impair a person's ability to talk, blink, swallow saliva, eat, or express themselves naturally with their faces. To detect this facial paralysis, the best technique is facial key point analysis. Even though it is the best technique, there are several limitations and drawbacks of facial key point analysis, including limited diversity, sensitivity to lighting, occlusion, pose variability, real-time performance, privacy concerns, etc. This paper proposes a method to overcome the limitation of light sensitivity, making it possible for facial key point analysis to detect key points on the face. In this paper, a composite model is implemented using a "generative adversarial network" (GAN) and "facial key point analysis". GAN is implemented to make the facial picture into a high-resolution picture. This GAN output is given for the input of the facial key point analysis. Facial key point analysis is the process of identifying and tracking specific points on a person's face, such as the corners of the mouth, the tip of the nose, and the eyebrows, to understand the movements and expressions of the face. The results of this composite model help detect facial paralysis more efficiently and accurately than before.