SVM and Logistic Regression for Facial Palsy Detection Utilizing Facial Landmark Features

Abstract

Facial Palsy is a problem related to temporary or permanent damage of facial nerve. Conventional technique for facial paralysis is physical detection and manual measurement for reconstruction of facial features in order to provide perfect balance of patient’s face. These Conventional techniques need to be strengthen using computational process. The present research work is carried out in this same direction. Facial palsy data collection and in continuation landmark coordination generation are challenging task. Landmark coordination is an input for learning model. Two machine learning models – Support Vector Machine and Logistic Regression are applied and these machine learning models will train the system using generated facial landmark features. The two important tasks for handling the facial palsy detection using machine learning are Landmark feature generation and effective machine learning model training. The outcome for facial palsy detection using support vector machine is better than logistic regression. The average accuracy achieved by support vector machine is 76.87%