Abstract

The present problem in analysing synchronisation is the human labour required. My objective is to simplify real-time synchrony recognition by extracting and analysing key synchronisation properties. My proposed method, computes synchronisation score for various parameters. To do so, I utilised OpenFace to analyse video footage of dyads conversing and used the bag of words approach to find structural similarities. A synchronisation score was generated utilising structural similarity and a concept of value proximity between the two signals. I subsequently looked at the results of various algorithm settings to better understand each parameter's impact. Finally, my technique was compared to other published algorithms and synchrony labelling conducted as part of a user survey. The results showed that my technique outperforms existing algorithms and is less vulnerable to noise. A user study also confirmed that the algorithm creates an intuitive synchrony depiction.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]