Ear Movement using BOOFCV

[BrandMetis Width]

Hello, 

We're working with a GP (Doctor) to create an Otoscope that can detect inner ear movement (Ear Rumbling) - using an Otoscope. 

We can do this - and have shown it can work via basic movement detection - but are now trying to make it more robust. 

We're looking to use BOOFCV's - KLT method to then apply motion detection on it (whilst using USB - OTG to keep it compatible with COTS cameras / Otoscopes. 

This seems to allow us to see whole movement (camera movement) vs clustered movement of the KLT - which gives us a good impression / probability of Typani motion. 

I wonder if anyone could suggest any methods for this last bit? To then look for clustered KLT changes in an image? It's the clustered movement that appeals to this - since there are lots of in-ear variables - e.g. light, cleanliness, narrowness of aperture - etc. 

Thanks 

John

[BrandMetis Width]

Is there any additional documentation of explanation of this feature:  

Pyramid Kit Tracker?
https://javadoc.io/static/org.boofcv/boofcv/0.13/boofcv/alg/tracker/klt/PyramidKltTracker.html