That's... five fingers, 3 joints each... 15 servos and 15 potentiometers (or similar joint-bend sensors).
Have you looked at your hand?
The vast majority of people -can't- bend their fingers at each joint independently; the control isn't evolved that way. Certain fingers can't even be bent independently without causing some movement of other fingers. YMMV, but this is generally how it is.
Now, it is possible to bend the second/third joints independently of the first joint on the fingers/thumb...
That effectively reduces your needs to 10 bend sensors (and really, you're going to want to use a bend sensor, not servos - why make it difficult on yourself?), but you could get away with 5. There is also what is known as "abduction", the side-to-side movement of the fingers and thumb; depending on the level of sensing needed, you could get away with as little as a single bend sensor between each finger, plus a couple on the thumb (located carefully).
What you're going to find, though, whatever method you use, is that the glove material is going to slide around. Good luck with that, it has plagued wired glove engineers for over 20 years (you have researched wired gloves, right?).
Your best best is a lycra glove, and to use optical bend sensors, which are essentially a fiber optic loop (sans jacket, and scored with marks to let out light as the fiber is bent) with an IR LED at one end, and a phototransistor at the other, connected to an amplifier, that can sense the bending of the joint via the light loss out the score marks (which reduce internal reflection). The phototransistor/amplifier is set up so that the output is a varying voltage signal, rather than a simple on/off signal (ie, the base of the phototransistor isn't driven to saturation, but acts more like one would in an audio amplifier).
This is the basics of a wired glove; essentially the idea and workings behind the VPL DataGlove (I am pretty sure the patents are expired now), sans some complex software (and this doesn't address position tracking of the hand/arm, which was originally done via expensive magnetic 3D tracking). There are other methods to sense the bending (resistive strain gauges, like used in the Mattel/Nintendo PowerGlove, as well as hall-effect sensors as used by another "exoskeleton" glove, IIRC, at CMU), but the lycra glove and optical bend sensors are a lightweight and fairly simple solution that is easy to put on and take off.
Even so, you are still going to be faced with the issue, over time, of the glove material shifting on the fingers and hand, necessitating their repositioning and recalibration. This may or may not be an issue for you in your application.
A possibility to consider, though, is using a PC and a high-speed USB web camera to track the position of the fingertips and hand using video; if you combined two or more camera outputs, you could potentially track everything in 3D (you need more than 2 cameras to do this effectively, because of potential occlusion issues). The hardware is the easy part, though (you would put IR LEDs or IR reflectors with an external IR source on the fingertips) - its the software that would be the killer. Tracking ten independent "blobs" of light in 3-space using a computer and few web cameras to intergrate the data isn't anywhere near as easy as it sounds. OpenCVS or other vision software and/or vision processing techniques will be paramount to success. It can be done, though, since there are plenty of (expensive) optical full-body motion capture systems out there used by companies for a variety of applications.
Good luck with your task, but you honestly have a lot of research and hard work to do. It is an area that could pay off fairly well if you see it to the end, though!
