[profinger]

Well the idea would be to get an average or something a "baseline" to filter leaning/talking animatedly/etc. Obviously it's probably not how it's done I'm just speculating. Obviously the coin difference would be something but that is something that can be within the margin of error. You give it a ~4lb window or something and all of that's covered. 10 quarters in a pocket is .12lbs.

Besides all of that, we're talking about within a few hours here not day to day. I just mean to give the person that beginners luck then drive them to keep playing. Obviously the accuracy wouldn't be all that vital but it could save the casino a lot of money if they're keeping track of that kind of stuff even if they only have a 50% accuracy.

[kbenson]

Keep in mind leaning on a counter or table, putting a leg up, eating and drinking. In the end, you are also putting weight sensors, which are fairly notorious for being fiddly, in a bunch of chairs. The wider you have to make the margin or error, the more false positives you'll get. I imagine an a medium size casino there's a lot of people that weigh ~180 lbs. In the end, it's a passive tracking system, where you get data events and then try to classify them, which I think is more error prone than actively tracking the people themselves through other means. With as many cameras as they have, it probably easier (at this point) to just make sure they always overlap, and have a motion tracking algorithm track people across the video feeds.

[zyxley]

If you're saying to use things like "140.2 pounds plus or minus 4 pounds" to try and differentiate among the many, many people in a casino... no, that's not really going to work.