Recognition of Human interactions is a problem as any other in Computer Vision. You should concentrate on feature selection, i.e. trying to find out those features which remains stable all over an action. So, if in your representation different actions give rise to different feature values your system will work well.

A good way to start is a small dataset, a bunch of good features, a check with PCA, and a simple classifier. It's enough to understand if the system you are thinking is promising or not.

I think that there is not the best method to detect human interactions..

There are a lot of methods, each one devoted to a particular application field and having its pros and its cons.. However, maybe that these two surveys can help you:

- Human activity analysis: A review, Aggarwal and Ryoo, Journal ACM Computing Surveys (CSUR), 2011

- Machine Recognition of Human Activities: A Survey, P. Turaga, R. Chellappa, V. S. Subrahmanian, and O. Udrea. 2008. IEEE Trans. Cir. and Sys. for Video Technol.

Regards, Alessia

Check this Eye tracking software 'Tobii' which can be useful for various environments.

http://www.tobii.com/en/eye-tracking-research/global/research/usability/

Thank you so much :)

Recognition of Human interactions is a problem as any other in Computer Vision. You should concentrate on feature selection, i.e. trying to find out those features which remains stable all over an action. So, if in your representation different actions give rise to different feature values your system will work well.

A good way to start is a small dataset, a bunch of good features, a check with PCA, and a simple classifier. It's enough to understand if the system you are thinking is promising or not.

there are a lot of algorithms used in detecting human gestures, and therefore guessing the state of the human. But i think, those algorithms still need to be improved and tested.

Ismail you are right but the success depends on how many actions you have. We reach about 75% in case of 7 actions. Is that a low rate? Surely to improve but yet usa le on some applications. Another little point: actions are somehow more general than gestures . The latter are currenlty better recognized.

Of course this is my experience.

Apart form the interesting previous responses, I recommend you to read our recently published paper about human tracking and detection of suspicious behaviors in shopping malls. Here, you can find some useful ideas about detecting human interactions in video-surveillance applications:

http://dx.doi.org/10.1016/j.eswa.2015.06.016

Good luck with your research!

Article Expert Video-Surveillance System for Real-Time Detection of ...