Detailed Activity Recognition from a Smartphone User's Position

I developed a detailed activity recognition method from a smartphone user's position (GPS, marked locations) with a Hierarchical Hidden Markov Model. Particle Filtering was used for the approximate inference of latent variables (activities). This method infers over twenty activities with high accuracy. Simulated in MATLAB. This was my research topic for my CSE undergraduate thesis.

  • Publication: CSE Thesis (pdf)

    Concentration: Sep. 2013 - Dec. 2013