Analysis of Markov Logic Networks for Activity Recognition

In this work we propose a method to recognise daily routines from combination of discrete activity patterns using Markov logic networks.

The use of Markov logic networks (MLN) enables us to combine first-order logic and probabilistic graphical models in a single representation. We use our HMM system to analyse and compare the performance of the two systems, we then report an analysis of MLNs for this experiment and show the experimental results that show the ability of our approach to model and recognise daily routines without user annotation. The results showed HMMs perform better in both accuracy and in computation speed but, MLNs may yet prove useful in future under different conditions.

The paper can be found here.

All the files can be found here here and here here.