Mobile sensors based platform for COVID-19 contact tracing leveraging artificial intelligence

COVID-19 pandemic is an active epidemic disease and is evolving in various variants of SAR-COV-2. The most critical point for COVID-19 is to break the transmission chain by early detection and isolation. It necessitates monitoring human interactions using remote technologies as a proactive strategy to minimize the spread. This paper presents a framework for user behaviour recognition and localization in a heterogeneous environment leveraging artificial intelligence for COVID-19 contact tracing. This approach designed and developed a mobile application to acquire the mobile sensors (received signal strength indicator (RSSI), accelerometer, gyroscope, and speed), sensors response, and machine learning-based user behaviour classification to assess the spread of COVID-19 with better proximity detection. This paper presents a machine learning-based free-space path loss (FSPL) model to estimate the distance between users holding mobile devices. The framework considers comprehensive user interaction scenarios depending upon multiple parameters, including the randomness of a user's interaction (device-device communication), device position, environmental obstacles, and device type. This platform classifies user behaviour on a train, bus, walking, and sitting at one, two, and three-meter from the suspected patient. The proposed system outperforms existing models, estimates the distance up to three meters, and accurately classifies each user behaviour. This system can send alert notifications to all users within three meters of a person whose COVID-19 test is positive. This approach uses machine learning models and finds that the random forest classifier shows higher accuracy of 97.95% for user behaviour classification and uses the LSTM model for distance estimation from one to three meters between users with the highest accuracy of 98.07%.