Transfer Learning in Sensor-Based Human Activity Recognition: A Survey of State-of-the-Art Research

Strategic Insights -

Researchers from the Georgia Institute of Technology have conducted a comprehensive survey of transfer learning methods in the application domains of smart home and wearables-based human activity recognition (HAR). The study, which covered 246 papers, highlighted the gaps in the literature and provided a roadmap for addressing these. The research aimed to summarize the existing works and provide a promising research agenda for the HAR community. According to the study, transfer learning has been explored extensively in this field, but large quantities of annotated data are typically not available for sensor-based HAR.

Key Takeaways:

  • The study focused on transfer learning methods in the application domains of smart home and wearables-based HAR.
  • The researchers analyzed 246 papers on transfer learning in sensor-based HAR and highlighted the gaps in the literature.
  • The study provided a problem-solution perspective by categorizing and presenting the works in terms of their contributions and the challenges they address.
  • The research aimed to summarize the existing works and provide a promising research agenda for the HAR community.
  • The study emphasized the importance of large quantities of annotated data for sensor-based HAR, which is typically not available.
  • Transfer learning has been explored extensively in this field, but more work is needed to address the challenges of real-world settings.
  • The study suggested a roadmap for addressing the gaps in the literature and providing a promising research agenda.
  • The research was conducted by Sourish Gunesh Dhekane, a researcher at the Georgia Institute of Technology.
  • The study was supported by financial support from CISCO.
  • The research aimed to provide a reference to the HAR community by summarizing the existing works and providing a promising research agenda.

Statistics:

  • 246 papers were analyzed in the study on transfer learning in sensor-based HAR.
  • The study highlighted the gaps in the literature, which can be addressed by providing a roadmap for forthcoming research.
  • The research aimed to summarize the existing works and provide a promising research agenda for the HAR community.
  • The study emphasized the importance of large quantities of annotated data for sensor-based HAR, which is typically not available.
  • Transfer learning has been explored extensively in this field, but more work is needed to address the challenges of real-world settings.

Sources:

  • "Transfer Learning In Sensor-based Human Activity Recognition: a Survey" by Sourish Gunesh Dhekane et al., ACM Computing Surveys, 2025;57(8):1-39.
  • Georgia Institute of Technology, College of Computers, Atlanta, GA 30332, United States (contact Sourish Gunesh Dhekane).
  • Assoc Computing Machinery, 1601 Broadway, 10TH Floor, New York, NY, USA (contact ACM Computing Surveys).
  • CISCO (financial support for the research).