Preliminary Diagnosis for Flu using Facial Feature Recognition and Thermal Camera

ABSTRACT

COVID-19 signs are similar to flu and their symptoms can range from no signs and symptoms (asymptomatic) to severe signs and symptoms (symptomatic). Fever, cough, fatigue, runny or stuffy nose, body aches, conjunctivitis, and headache are some of the symptoms that COVID-19 and flu have in common. Current technologies only take advantage of thermal sensors to identify the individual with possible flu-like symptoms. Adding extra layer of security in preventing the spread of diseases for people with flu-like symptoms must be set up at key locations using a video camera. The study can detect facial features such as reddish eyes, runny nose, dark circles around eyes and measure temperature using a non-contact thermal camera. Gathered image datasets were used for model training. Initial testing with the system revealed that closer distance and better illumination yielded better results upon consultation with a doctor using the comparison of 100 LUX and 1000 LUX lighting conditions. Validations were done using live video feeds where PCA and SVM were used for feature extraction and classification respectively. Support Vector Machine was used to evaluate subject whether they exhibited flu-like symptoms or not and compare the system output with doctors diagnosis. Error rates of 26.67% and 50% were achieved for False Acceptance and False Rejection Rates respectively along with 0% error rates for the temperature detection system. © 2021 IEEE.