Date: Wednesday, March 26, 2025.
Time: 12 PM – 2 PM EDT.
Location: C1215 Midtown, CODA and Zoom meeting (ID: 931 754 7921 Passcode: 042839)
Shruthi K. Hiremath
Ph.D. Candidate in Computer Science
School of Interactive Computing
Georgia Institute of Technology
Committee:
-----------------
Dr. Thomas Ploetz (Advisor), School of Interactive Computing, Georgia Institute of Technology
Dr. Gregory Abowd, School of Interactive Computing, Georgia Institute of Technology
Dr. Sonia Chernova, School of Interactive Computing, Georgia Institute of Technology
Dr. Diane Cook, School of Electrical Engineering and Computer Science, Washington State University
Dr. Uichin Lee, School of Electrical and Computer Engineering, Korea Advanced Institute of Science and Technology
Abstract:
-----------------
Smart Homes have come a long way: From research laboratories in the early days, through periods of (almost) neglect, to their recent revival in real-world environments enabled by the existence of commodity devices and robust, standardized software frameworks. With such availability, human activity recognition (HAR) in smart homes has become attractive for many real-world applications, especially in the domain of Ambient Assisted Living (AAL).
Yet, building an activity recognition system for specific smart homes, which are specialized spaces with varying home layouts and inhabited by individuals with idiosyncratic behaviors and habits, is a non-trivial endeavor. For real-world deployments, privacy and logistical concerns essentially rule out the possibility of third parties being able to collect the much-needed annotated sensor data while the resident already lives in their smart home.
I address the challenges of developing a Human Activity Recognition (HAR) system for smart homes by defining its lifespan of three phases: i) bootstrapping, ii) updating and extending, and iii) expanding recognition capabilities for complex tasks. In the bootstrapping phase, I establish a system that quickly recognizes prominent activities with minimal resident involvement. The second phase introduces an update and extension procedure to improve segmentation accuracy for the activities previously identified. Finally, I enhance the system's capabilities by incorporating large language models (LLMs) as contextual knowledge bases. LLMs help encode contextual information through language-based descriptions and identify structural constructs of complex activity sequences, aimed at improving recognition and monitoring changes in routine patterns.
]]>