I believe this product of NDI measurements sciences (link) is one of the best on the market, because the movement of the fingers can hide from cameras, then this analysis using electromagnetism becomes quite efficient.
If you want to capture human joints motion I would not recommend a data glove as it will nail you to the specific kinematic model designed by the manufacturer of the capture system. Data gloves are fine if you just want to record data without worrying about which joints motion have been captured.
If you want to to model a human hand kinematics yourself, I would also question the use of a vision system, such as vicon (one of the best on the market anyway). Such systems are vulnerable to occlusions that might happen when you rotate the wrist, also markers might be difficult to stick on the body (the vicon one at least). A motion tracker can do the job well, in my opinion, as it is just following the bone displacement at the cost of making the subject a bit uncomfortable. If well designed, the uncomfort should not damage the motion serverely enough in order not to be natural.
NDI are good ones but fragile and a bit pricy, you might want to see also how the scientist of this project done: http://grasp.xief.net/
Please consider looking at our active marker based mocap system. There are lots of benefits over alternative systems including:
Small markers that can be attached to digit segments.
Markers have individual identity (which will save you lots of post processing time)
System has very high signal to noise - high accuracy (LED emitters versus reflectors)
You don't need lots of cameras - one CX1 scanner can potentially do the job of several cameras.
Very wide field of view - you get a large measurement volume from a small set-up.
Portable and can be used outdoors or under untreated lab lighting.
Price - Our systems are modular, so you can start with a small set-up and build to increase you measurement volume or to work win more complex environments..