PhD Dissertation Proposal Defense: Bhawana Chhaglani, Privacy-Aware Ubiquitous Sensing Systems for Healthy Indoor Environments
Content
Speaker
Abstract
With the majority of our lives spent indoors, it is imperative to maintain healthy indoor air quality for our well-being. Poor indoor air quality directly impacts our health and productivity, leading to long-term negative effects. The importance of healthy air is further heightened during pandemic or flu season or for immuno-compromised individuals at higher risk of contracting airborne transmission. While smart building systems focus on optimizing comfort and energy efficiency, techniques that address safety and wellness—particularly in mitigating airborne transmission risks remain underdeveloped. Estimating these risks requires complex, multi-faceted information about ventilation, aerosol generation, and occupancy. The existing solutions to sense these parameters are often invasive, expensive, or impractical for widespread use. Addressing this gap, pervasive sensing techniques present an opportunity to non-invasively monitor the whole spectrum of indoor air quality information, which can have a significant effect on an individual's health.
In this thesis proposal, I will present novel, privacy-preserving sensing systems designed to monitor ventilation rate, aerosol emissions, and crowd dynamics, empowering users with critical information to make better decisions while assisting the organization stakeholder in maintaining standards. I leverage ubiquitous sensing modalities such as microphones to provide deployable, non-invasive solutions for healthy indoor environments. By focusing on privacy-aware techniques and robust sensing methodologies, these systems offer scalable solutions for monitoring air quality, assessing transmission risks, and modeling crowd dynamics.
First, I will present FlowSense, a non-invasive method for monitoring building ventilation using audio signals. By capturing low-frequency airflow sounds, FlowSense predicts vent status and airflow rates with high accuracy, leveraging machine learning models and privacy-preserving signal processing techniques. I show that FlowSense enables users to assess ventilation adequacy in real time, promoting healthier indoor air quality and improved decision-making. Next, I will describe AeroSense, an ambient acoustic sensing system designed to predict aerosol emissions from human activities. Aerosol is a key indicator of airborne transmission risk which has received less attention despite its importance. By using non-reconstructible audio features, AeroSense detects and localizes everyday activities such as speaking, coughing, or sneezing, as well as mask usage and interpersonal distances. By combining these parameters, the system predicts aerosol levels, helping users and building management optimize ventilation strategies. Extensive evaluations in both controlled and real-world settings demonstrate AeroSense's effectiveness in mitigating transmission risks while maintaining user privacy.
To address broader privacy concerns in audio sensing, I propose FeatureSense which explores generalizable privacy-aware audio features that reduce speech and speaker-related privacy leakage while maintaining the effectiveness of audio-based applications. By designing a comprehensive privacy evaluation framework and adaptive task-specific feature selection, this work provides a foundational framework for ensuring trust in audio sensing technologies. Finally, I propose CrowdSense, which leverages passive WiFi logs to predict crowd densities and provide personalized alerts to users based on their trajectories. By integrating spatial, temporal, and semantic data, CrowdSense offers users actionable insights into crowded areas, enabling proactive decision-making to avoid high-density spaces. This event-driven approach can help reduce health risks associated with overcrowding, particularly during disease outbreaks or other high-risk scenarios. Collectively, these systems pave the way for safer, smarter, and sustainable indoor spaces while ensuring privacy.
Advisor
Prashant Shenoy