Hardware-accelerated machine learning for human localization, identification and behavior inference

Use energy-efficient machine learning techniques, including  (recursive) neural networks, for tagless indoor human  localization, identification and behavior inference using data  from multiple sensor types.  For high energy efficiency, data should be processed using  hardware accelerators and close to source, e.g., using embedded  low power FPGAs.  Behavioral inference can much benefit from (accelerated) cloud  processing, using self-correlation and cross-user correlation of  large amounts of data.

Link:

ERC Sector:

  • PE7_7 Signal processing
  • PE6_11 Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video)

Keywords:

  • Machine learning
  • Low-power electronics
  • Embedded machine learning

Research groups