Recognition Issues Between B and D, C and E

[whitedavidp]

This is really working out pretty well for me. Thanks!

I am noting, however, ongoing confusion between the following alpha characters:

B and D - with D prevailing far too frequently

C and E - with C prevailing far too frequently

I am at a loss to explain this result. I have loaded the raw resource into the Gesture Builder app and it has no problem distinguishing the correct characters when testing - correct virtually every single time.

Now I am sure that the recognition algorithm may be different. I have briefly looked at the code for both apps. I see in this app weighted libraries that I do not think I see in the Gesture Builder code (I could be wrong though). So I wonder if that has anything to do with it?

Best

[whitedavidp]

Actually, now that I think about this and stretch my aging memory, I recall modifying the original GestureBuilder app to use PointCloud recognition rather than the more common mechanism. So my earlier comment about GestureBuilder doing better is suspicious. I need to go back and find my code somewhere. If this is the case, perhaps now that I have forked, I can try to retro-fit PointCloud into this app. Best!

[tmatz]

That's interesting. I'll keep an eye on the results.

[whitedavidp]

At this point, in my fork, I have had to entirely replace the gestures for "b" and "e" to avoid the frustrating, frequent mis-recognitions.

A better understanding on my part of how the original works will be needed before I can try to retro-fit PointCloud recognition (if that proves needed at all).