It's slow, but you can use the Matlab floating point toolbox. Created a fixed point container and specify the bit-width and fraction length
a = fi(<float_value>, 1, <bit_width>, <fraction_length>);
The 1 denotes a signed fixed-point number.
"a" now contains your floating point number as a fixed-point. Of course, depending on your bit_width and fraction_length, you may lose some resolution. Now use "a.bin" which returns the binary string of the fixed-point container.
At that point, just write "a.bin" out to your file.
Converting from binary to float is a little more complicated. If someone has a more direct method, please let me know.
a = mvl2dec(<binary_string>);
mvl2dec converts a multi-valued logic string to a decimal value. "a" now contains a decimal representation of your number. Now since it is in fixed-point representation, divide a by 2^<fraction_length>. This will return your float. So you could do it in one shot with
a = mvl2dec(<binary_string>) / 2^<fraction_length>;