Privacy Preserving Individual Activity Recognition with Wireless Signals

Activity recognition is of great importance for various applications, such as Augmented/Virtual Reality, Sports Tracking and Security. Common automated alternatives require the capturing of images which may compromise the privacy of those being monitored, therefore an accurate alternative is required. The main goal of this thesis is: using Channel State Information [1] from Wi-Fi signals, build, test and evaluate an accurate activity recognition system. An existing example can be found here [2].

Your tasks will be:

  • Build a robust system to collect CSI values from Wi-Fi signals, although previous work at our group is already available from which your work should be based on.
  • Collect a large dataset of various activities which must be accurately detected with the use of advanced and robust machine learning methods.
  • Evaluate the accuracy of your approach against alternative methods, e.g., using accelerometers or any other sensors.

Requirements

  • Good experiences with Python or R
  • Solid and extensive knowledge of C
  • Familiarity with Machine Learning

Good to have

  • Good knowledge about wireless signal properties
  • Experience with deep learning and signal processing

If you are interested, please send me an email with your CV and transcripts and a short introduction.

Leonardo Tonetto --- tonetto at in tum de

[1] https://en.wikipedia.org/wiki/Channel_state_information

[2] github.com/linteresa/WiAR