It sounds like it is a quasi-IEEE (in other words, not a standard float or double, but a representation that has the same form only with the number of bits different).

Anyway, I can't really help you unless you have the formula to convert the representation to a decimal number (or a number in another base).

The only other suggestion I have is to see if its a fixed point representation. Fixed point is just a way of storing a number with so many possibilities for the integer value and so many possibilities for the non-integer value. Based on what you have said about the structure of the number, I don't think its a fixed point representation.

Take a look at this and see if it has a connection:

IEEE floating point - Wikipedia, the free encyclopedia