How to read LSB of audio sample?

Hello everyone... I am working on audio steganography project.And i need to read and manipulate LSB of the audio sample.Please tell me a way to do this. Thanks in advance.

[manisha sharma]

The audio samples are in floating point representation.When i use dec2bin to convert the floating point number in binary, it only shows binary equivalent of digits before decimal point.But i want to reach LSB of the sample.

[James Tursa]

Do you simply want to overwrite the trailing least significant byte of the mantissa with a different bit pattern? I.e., change the floating point value slightly but have your desired pattern embedded in the trailing bits?

Answer 1

[James Tursa]

Here is an example of overwriting the least significant byte of a double with other bit patterns, in this case the string 'hidden':

% Arbitraty 6 element double vector
>> x = rand(1,6)
x =
   0.957166948242946   0.485375648722841   0.800280468888800   0.141886338627215   0.421761282626275   0.915735525189067

% Find the location of the least significant byte in the double for this machine
>> LSB = find(typecast(1,'uint8')~=typecast(1+eps,'uint8'))
LSB =
     1

% Re-interpret the bytes of the double vector as a uint8 vector
>> u = typecast(x,'uint8')
u =
  Columns 1 through 35
  153  127  112  148   28  161  238   63   82  133   98    6  101   16  223   63  124   40   48  201  229  155  233   63  216  151   19  224   84   41  194   63  202   83  230
  Columns 36 through 48
    8   35  254  218   63   32   30  143  150  180   77  237   63

% Our replacement data
>> s = 'hidden'
s =
hidden

% Construct the indexes for the replacement bytes
>> lower_bound = LSB
lower_bound =
     1
>> upper_bound = LSB + 8*(numel(s)-1)
upper_bound =
    41

% Replace the floating data with our new data
>> u(lower_bound:8:upper_bound) = s
u =
  Columns 1 through 35
  104  127  112  148   28  161  238   63  105  133   98    6  101   16  223   63  100   40   48  201  229  155  233   63  100  151   19  224   84   41  194   63  101   83  230
  Columns 36 through 48
    8   35  254  218   63  110   30  143  150  180   77  237   63

% Turn result back into double vector (note result is close to original x)
>> y = typecast(u,'double')
y =
   0.957166948242940   0.485375648722843   0.800280468888797   0.141886338627212   0.421761282626269   0.915735525189076

% Now do the reverse to recover our hidden data

% Re-interpret the bytes of the double vector as a uint8 vector
>> w = typecast(y,'uint8')
w =
  Columns 1 through 35
  104  127  112  148   28  161  238   63  105  133   98    6  101   16  223   63  100   40   48  201  229  155  233   63  100  151   19  224   84   41  194   63  101   83  230
  Columns 36 through 48
    8   35  254  218   63  110   30  143  150  180   77  237   63

% Pick off our hidden bytes and re-interpret as a string
>> char(w(lower_bound:8:upper_bound))
ans =
hidden

[manisha sharma]

Thank you so much... this will be really helpful. Thanks once again.

[manisha sharma]

I have no knowledge about uint8.So can you please tell me about the use of uint8 (in the above example you have given).

Also it seems that for decoding we need to know the length of message 's'.Is it so?

Because while calculating Upper_bound we need length of 's' ie. our text message. But on the decoding time we have our encoded audio only.So i think, i need to embed the value of Upper_bound in audio itself.

This is what i think.Please can you clear my doubts?

Thanks...

[manisha sharma]

This method is not working for audio samples.May be because audio samples have large digits after decimal, example 3.0518e-004

What should i do?