Hi Cesar
I am working on the power disaggregation problem, where we want to find which appliances in a house are switched on or off given an observed total power signal, sampled at around one second intervals. Several papers suggest the use of a Factorial Hidden Markov Model (Ghahramani and Jordan 1996)
http://18.7.29.232/bitstream/handle/1721.1/7188/AIM-1561.pdf?sequence=2 .
Each component HMM in the Factorial HMM models a single appliance. The trick is to decode the observed power measurement sequence to get back to the individual appliance states.
I wondered if you had any thoughts on how best to construct a factorial HMM from the Accord.Net HMM class, to do the job.
Thanks for you consideration
Andy