Reproduce “Maintenance Required: Updating and Extending Bootstrapped Human Activity Recognition Systems for Smart Homes” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Shruthi K. Hiremath, Thomas Ploetz

A framework is presented for continuously updating and extending bootstrapped Human Activity Recognition (HAR) systems in smart homes by combining self-supervised learning with dynamic motif model updates. This approach consistently enhances the segmentation accuracy of prominent activities, adapting to evolving resident behaviors with minimal manual intervention.

Key insights

  • Continuous, data-incremental updates are crucial for maintaining and improving the accuracy of smart home HAR systems as resident behaviors evolve.
  • Self-supervised learning can effectively learn robust data representations from unlabeled ambient sensor data, significantly reducing the annotation burden for ongoing system adaptation.
  • Improving the precision of activity segmentation, rather than just classification, is essential for granular understanding and practical utility in ambient assisted living applications.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.