I am assuming that you have (starting from the left end) one sign bit, the assumed binary point, and then seven bits that represent a fractional value. If that's the case then you can just take the signed integer value of the fixed-point number and divide by 128. You'll need to do this division using floating-point values, of course, because the result will be less than 1.
The range of values that can be represented in this fixed-point format is -1.0 to +(127/128).