4

In MATLAB (r2009b) I have a uint32 variable containing the value 2147484101.

This number (its 4-bytes) has been extracted from a digital machine-vision camera in a grabbing process. According to what I understand it holds the single-precision form of the shutter-speed of the camera (should be close to 1/260s = 3.8ms).

How do I convert this 32-bit number to its IEEE single-precision floating-point representation - using what's available in MATLAB?

With mentioned value in variable n, I have tried using a combination of nn=dec2hex(n,16) and then hex2num(nn). But it seems that hex2num expects the hexadecimal coding to be double-precision and not single as it is here. Atleast I am getting weird numbers with this method.

Any ideas?

Edit: Tried @Matt's answer below:

typecast(uint32(2147484101),'single') %# without swapbytes
typecast(swapbytes(uint32(2147484101)),'single') %# with swapbytes

Which gives:

ans =

  -6.3478820e-043

ans =

  -2.0640313e+003

I tried the IEEE 754 converter (JAVA applet) at http://www.h-schmidt.net/FloatApplet/IEEE754.html.

Using:

format hex
typecast(uint32(2147484101),'uint8') %# without swapbytes
typecast(swapbytes(uint32(2147484101)),'uint8') %# with swapbytes

gives

ans =

   c5   01   00   80

ans =

   80   00   01   c5

Entering these bytes into the applet (hexadecimal) gives me the same numbers as MATLAB.

Amro
  • 123,847
  • 25
  • 243
  • 454
Ole Thomsen Buus
  • 1,333
  • 1
  • 9
  • 24

2 Answers2

8

I think what you're saying is that the underlying bits represent a floating point number, but that you've got it stored as a uint32.

If that's that case, you can cast it (i.e. reinterpret the bits) as a single precision float using the typecast() function.

b = typecast(a, 'single')

where a is your variable.

See: http://www.mathworks.com/help/techdoc/ref/typecast.html

Edited: not the cast function, the typecast function... My apologies!

Matt
  • 425
  • 5
  • 11
  • Yep, it must be `typecast` and not `cast`. But I am not really getting values I hoped for. You see, I have no prior-knowledge about the numbers. I am not even sure I got them stored correctly :) I just cross-checked using a java applet. So your method is sound. +1 and accepted. – Ole Thomsen Buus Feb 15 '11 at 11:34
  • @Ole Thomsen Buus: It's hard to suggest what you might do differently without knowing the method by which you arrived at your uint32, or the specification of the file/device you're reading. Hopefully you'll figure out where it's going wrong... – Matt Feb 15 '11 at 15:42
2

You could do the cast when you read the data in with fread().

Have a look for the precision argument, you could read it as the int32 number and store it as a single by doing 

shut_speed=fread(fid,1,'int32=>single');
mor22
  • 1,372
  • 1
  • 15
  • 32