kentaylor / AvocadoAC

Inspired by Chris Smith's project this application does classification differently, logs classifications in a local database and maintains an online repository of activities.
https://github.com/kentaylor/ContextApi/wiki/Classification-Update
5 stars 5 forks source link

Off Algorithm For My Tracks #67

Open kentaylor opened 13 years ago

kentaylor commented 13 years ago

Off algorithm for My Tracks needs improvement. The ideal algorithm for starting and stopping My Tracks is very dependent on the classifier. For example we do not want My Tracks to start when active inside the house but do want it start quickly when going on a walk. Perhaps we could further refine walking to detect purposeful walking versus walking around the house by requiring the walk rate to be constant through a sampling interval or/and requiring the same walking rate to exist across multiple sampling intervals. FFT would be useful for this algorithm. Perhaps we have a new activity walking vs purposeful walking.

Currently the starting algorithm is not too bad but My Tracks is running a long time after walking stops. This is because a single walking event will reset the off timer for another 5 minutes. An algorithm to counter this would be, once a non exercise activity is detected then stop My Tracks if the exercise activity is not detected for 2 out of the next 3 minutes.