SparseSense: Human Activity Recognition from Highly Sparse Sensor Data-streams Using Set-based Neural Networks

  • 2019-06-06 03:50:33
  • Alireza Abedin, S. Hamid Rezatofighi, Qinfeng Shi, Damith C. Ranasinghe
  • 18

Abstract

Batteryless or so called passive wearables are providing new and innovativemethods for human activity recognition (HAR), especially in healthcareapplications for older people. Passive sensors are low cost, lightweight,unobtrusive and desirably disposable; attractive attributes for healthcareapplications in hospitals and nursing homes. Despite the compellingpropositions for sensing applications, the data streams from these sensors arecharacterised by high sparsity---the time intervals between sensor readings areirregular while the number of readings per unit time are often limited. In thispaper, we rigorously explore the problem of learning activity recognitionmodels from temporally sparse data. We describe how to learn directly fromsparse data using a deep learning paradigm in an end-to-end manner. Wedemonstrate significant classification performance improvements on real-worldpassive sensor datasets from older people over the state-of-the-art deeplearning human activity recognition models. Further, we provide insights intothe model's behaviour through complementary experiments on a benchmark datasetand visualisation of the learned activity feature spaces.

 

Quick Read (beta)

loading the full paper ...