T thought this seemed interesting, and not that expensive if it is as good as it looks, and can be used with blender.
I posted about this well over a year ago IIRC.
I contacted Qumarion about the pricing, availability and the SDK/documentation as I wanted to implement it for Blender.
I never got any answer to my inquiries - screw them. It’s likely another product that’ll hardly leave their island 
or perhaps they don’t have anyone who can do English PR? It’s not surprising to get no reply email from from a Japanese company because of a language barrier.
Nope, they have (or had, don’t care anymore) a designated english form. And I frequently got something to do with japanese companies, many have very good english although “only” a secretary or a random guy is replying. There’s usually some japanglish in it, but that’s a problem with the education. Japanese schools teach a lot of english, but only vocabulary and grammar grinding - thus they often have problems to have an actual conversation forming wonky sentences translated 1:1 from japanese or using japanglish stuff.
I even asked details as I wanted to import several boxes - not even trying to reply to me means they don’t want to do business.
I watched the video up to the “~$700 US” part then went, “meh…”
Which does answer the pricing question.
$750 US (at 2:00 in). Not bad for what it is. Not quite motion capture, but a lot more intuitive for animation. Hell, how many people here have paid close to that for a graphics card? Not me, but damn…I think if this worked with Blender, $750 would open the door to a lot of intuitive creativity. But yes, I doubt we’ll see this working with Blender in the near future.
Make your own? 
The parts are readily available, could even use lego mindstorms, arduino or the like. It might not be as small or pretty, but I’m sure it could be made to work as well.
Does a general interface exist for Blender to read arduino sensors?
[EDIT] I guess this approach could be used : http://shallowsky.com/blog/2011/Oct/16/
I had thought about making something similar to that for inputting poses for animation. While I still think that for going from default pose to a pose that is close to what you want it would be quite good, tweaking existing poses would be a pain unless it could move it’s joints to match an existing pose.
Currently taking a class on launching entrepreneurial ventures and I wanted to make a simplified version of this Qumarion for the US for a much lower price tag.
If anyone’s interested hit up my survey and say you’d love to buy 1000 of them and maybe the professor will endorse it going forward.
Bump I only got 1 response 
Really need at-least a few more lol :(:(:(
I’m very interested in this but prefer to open source it (maybe not).
Here is what I have in mind, modular joins.
There are 2 major use cases:
- mocap of human (for impossible poses)
- Non human, non biped (shortcut to keyframe animation)
If you have the skill, I’ll throw money your way.
I’ll consider it seriously. I need to read up on the implications of going open source.
Very interesting prospect though Light Bwk X ).