It's a good idea. Myoelectric sensing is currently much more reliable and accurate than camera-based sensing. The algorithms involved are much simpler and more robust. I can actually see myself using something like this, whereas current-gen CV-based equipment really isn't good for anything but novelty.
Anyone actually received theirs? I ordered and then upgraded to their DevKit Program, when the emails when out about that. The article showed a picture of a pretty production looking device.