Motivation:
Indoor air quality is very important when it comes to ensuring the safety and comfort of individuals in a variety of settings. However, the existing air quality monitoring systems are often costly and do not fully consider risks assoiated with respiratory droplets and aerosols. Therefore, a tool that can provide information regarding the duration of exposure to resporatory aerosols would prove to be incredibly beneficial. By utilizing such a tool, individuals and organizations would be better equipped to make informed decisions regarding the safety of indoor environments, thereby minimizing the risk of respiratory infections and promoting overall health and well-being.
Goals:
In this project, we aim to develop application to monitor resident time of human respiratory aerosols in indoor environment utilizing mobile sensors and machine learning models, with the aim of improving the safety of people, especially in high-risk environment such as hospitals, healthcare facilities, and classrooms. The app will capture syndromic signal such as cough and provide dinformation to optimize safety in dynamic real world setting and use cost effective and accessible to the general public.
Methods
The project is divideed into three stages: Data collection, Data Visualization, and Modeling and Simulation.
Data Collection
Data Collection APP Development
We developed an iOS application for collection of data and proof of concept deployment of models.
Our app includes features such as:
Audio: capture and classify audio to detect human respiratory events such as cough, sneeze
Thermal Image: using a FLIR one camera to detect surface temperatures, and detect human presence, and movement in the thermal image using YOLO model
Lidar and Camera: capture room layout info and geomtry required for modeling and CFD simulation
Database: store collected data online using Firebase
APP Content View
Tools for Measuring Aerosol Concentrations
We used the SPS30 Particulate Matter Sensor to measure our ground truth aerosol concentration.
Data Collection Process
We established a test environment in a compact office space (approx. dimenisions - 3.2m, 2.6m, 3.2m), where we mechanically simulated human coughs and captured aerosol concentrations using particulate matter (PM) sensors.
The cough simulation involved a mannequin, mechanical ventilator, fog machine, and air compressor.
We have installed six PM sensors to measure the actual particle concentration in the room. The location of each sensor is indicated in the layout.
The iPhone is mounted on a tripod in front of a mannequin to capture thermal images and audio. In the picture, three sensors can be seen: one directly in front of the mannequin, one positioned on the tripod with the iPhone, and another located on the side of the wall close to the door. These sensors are used to measure aerosol concentrations.
The fan machine is located in the right back corner, which we refer to as the exhaust area, and one sensor is positioned there to measure aerosol concentrations. Another sensor is placed on the wall behind the mannequin.
Our data collection app, which we have developed for this purpose, allows us to capture thermal images of the room using a thermal camera. In the image, the temperature of the mannequin is shown as 46.5 degrees Celsius, due to the heat generated by the fog machine.
Data visualization
Exploratory Data Analysis
Our exploratory data visualization has revealed that the dispersion and duration of aerosol concentrations at various sensor locations varies under different fan speed settings during a cough event.
Impact of Fan Speed and Sensor Location on Aerosol Concentrations
