A sensor to identify 3 different shapes

[Minu]

I am making a prosthetic hand for a competition. For that i have to solve a task with that hand. The task is to identify 3 shapes(square, cylinder and ball) with out seeing that. As per the rule i cannot use a camera. Is there any sensor to identify all these shapes. The sensor should be firmly attachable to the hand without disturbing the normal activities of the hand.

 

[Harald Kapp]

To me this sounds like a task for pressure sensitive sensors on the tips of the "fingers" of your "hand". I guess that's how blind people would tackle this problem. However, it is probably not going to be easy to conclude the form of the object from the orientation of the fingers and the response of the pressure sensors.
Another idea is to use ultrasound to detect the form, similar as an unltrasound imaging device doctors use to image interior soft parts of the body. This is kind of an acoustic camera. Would this be considered cheating with respect to the rules?

 

[Bluejets]

Two sensors in a clamp arrangement, bit like a set of calipers with a sensor on each side, probably at the hinge point.
Rotate either the clamp while holding the object firm or visa versa.
Rotation would have to be both latitude and longtitude.
Rotation would not have to be much at all, could be 2 x servo driven via a gimbal for example.

If a ball, no change.
If a square, change both ways.
If a cylinder, change 1 way.

Don't forget to mention you got the idea elsewhere in your homework.

 

[BobK]

Do you have prior access to these objects? If they are all different weights there is an easy cheat.

Bob

 

[Bluejets]

Do you have prior access to these objects? If they are all different weights there is an easy cheat.

Bob

I like the way you think outside the box.

 

[Nanren888]

Use the hand to scan a single photodiode across the scene?
.
A palm reader?
Multiple tactile sensors and a palm that goes flat.
.
What sort of hand construction, structure are you going for?
If you want tactile, on the grippers/fingers the gripper positions could be optimised for the shapes to be classified.