This Crosstalk Cancellation Function kills my computer!

Dont run this function. but please tell me how to fix it! it is a very basic cross talk cancellation filter. It takes a little while to compute, less than a minute, then it outputs a .wav file into the Current Folder Window. THEN, if you attempt to open that file, inside matlab, outside in iTunes, wherever. Your computer will DIE!!! System overload, have to do a complete shutdown. I dont think there is much heavy computation to this function, in fact other functions I've written have been more computationally complex on the CPU. So this is a mystery to me why it isn't working properly. Any ideas?

code:

% FUNCTION DESCRIPTION
%
% XTalker
%   function takes 2 channels of 3d surround audio and removes deleterious
%   crosstalk signal.  KEMAR HRTF impulse responses from 
%   http://sound.media.mit.edu/resources/KEMAR.html were used.  For optimal
%   results please listen through stereo speakers 2 to4 meters apart at 30 and 330 
%   degrees, 0 elevation, and 1.4 meters distance from your head. Also adds a 10 kHz
%   lowpass filter to widen the sweet spot.
%
% Inputs:
%       1)  INFile          : Name of stereophonic .wav file needing xtalk
%       2)  OUTFile         : Name of xtalked .wav file to which the  
%                             signal will be written
%
% Output:
%       A crosstalk cancelled .wav file of the INFile named OUTFile
%

function XTalker( INfile, OUTfile )

% BASIC ERROR CHECKING

if ischar(INfile) == 0,
   error('INfile input needs to be a string');
end

% number of arguments check.

if nargin ~= 2,
    error('number of inputs is incorrect')
end

% READ THE FILES

% The function reads the input .wav file
[y, Fs] = wavread(INfile);
RChannel = y(:,1);
LChannel = y(:,2);  

% Function makes sure the sampling rates match up.  (The sampling rates of 
% the files I provided and the KEMAR impulse responses line up at 44100.  
% If user chooses to use different 3D input file, and or different HRTF IRs,
% please make sure they are of the same sampling rate.
if Fs ~= 44100 
    error('HRTF IR and signal sampling rates dont match');
end

% Check length of the signal channels to make sure they are the same    
if length(LChannel) > length(RChannel);
    zeropad = length(LChannel) - length(RChannel);
    RChannel = [RChannel; zeros(zeropad, 1)];
end

if length(RChannel) > length(LChannel);
    zeropad = length(RChannel) - length(LChannel);
    LChannel = [LChannel; zeros(zeropad, 1)];
end

% Reads compact data file of + and - 30 degrees HRTF IRs from KEMAR.
  fp = fopen('L0e030a.dat','r','ieee-be');
  data = fread(fp, 256, 'short');
  fclose(fp);
  L2RIR = data(1:256);
    L2RIRLength = length(L2RIR);

      fp = fopen('L0e330a.dat','r','ieee-be');
    data = fread(fp, 256, 'short');
    fclose(fp);
    L2LIR = data(1:256);
      L2LIRLength = length(L2LIR);

      fp = fopen('R0e030a.dat','r','ieee-be');
    data = fread(fp, 256, 'short');
    fclose(fp);
    R2RIR = data(1:256);
      R2RIRLength = length(R2RIR);

      fp = fopen('R0e330a.dat','r','ieee-be');
    data = fread(fp, 256, 'short');
    fclose(fp);
    R2LIR = data(1:256);
      R2LIRLength = length(R2LIR);

% NORMALIZATION TO PREVENT CLIPPING AND MAKE STABLE FILTERS

% Find the maximum value of the samples in the stereo matrix
    maxLChannel = max(abs(LChannel));
    maxRChannel = max(abs(RChannel));
    maxVec = [ maxLChannel , maxRChannel ]';
    maxVal = max(maxVec);

% Normalize all samples in the stereo pair matrix to that maximum value sample.
    y = y / 1.0001 * maxVal;

% Find the maximum value of the HRTF IR data points
    L2RMax = max(abs(L2RIR));
    L2LMax = max(abs(L2LIR));
    R2LMax = max(abs(R2LIR));
    R2RMax = max(abs(R2RIR));
    maxiVec = [ L2RMax, L2LMax, R2LMax, R2RMax ]';
    maxiVal = max(maxiVec);

% Normalize all HRTF IRs to the maximum value
    L2RIR = L2RIR / 1.0001 * maxiVal;
    L2LIR = L2LIR / 1.0001 * maxiVal;
    R2LIR = R2LIR / 1.0001 * maxiVal;
    R2RIR = R2RIR / 1.0001 * maxiVal;

% CONVOLVE SIGNALS WITH HRTF IMPULSE RESPONSES TO CREATE THEORETICAL SIGNAL
% OF WHAT YOUR EARS WOULD HEAR IF THE SIGNAL WAS PLAYED TRANSAURALLY    
% convolve : take FFTs, multiply, take iffts
    LC2RE = conv(L2RIR, LChannel); % alternate side CROSSTALK SIGNAL
    LC2LE = conv(L2LIR, LChannel); % same side
    RC2LE = conv(R2LIR, RChannel); % alternate side CROSSTALK SIGNAL
    RC2RE = conv(R2RIR, RChannel); % same side

% CREATE ANTI PHASE SIGNALS OF CROSSTALK SIGNALS 
    XLC2RE = LC2RE * -1;
    XRC2LE = RC2LE * -1;

% NOTE: ACCORDING TO THE KEMAR WEBSITE THE INTERAURAL TIME DELAY IS STILL 
% PRESENT IN THE TWO EARS, SO THERE IS NO NEED TO ACCOUNT FOR THIS 
% INFORMATION WHEN SENDING THE SIGNAL

% ADD ANTI PHASED CROSSTALK SIGNALS TO ALTERNATE CHANNELS
% find length of INFile channels
    lengthCHANS = length(LChannel);
% find length of Convolved Signals
    lengthCONVS = length(XLC2RE);
% zeropad channels
    zeropad = lengthCONVS - lengthCHANS;
    LChannel = [LChannel; zeros(zeropad, 1)]; 
    RChannel = [RChannel; zeros(zeropad, 1)];   
% Add signals
    XLChannel = LChannel + XRC2LE;
    XRChannel = RChannel + XLC2RE;

% ADD 10 kHz BUTTERWORTH FILTER TO WIDEN THE SWEET SPOT
    Wn = 10000 / (Fs/2);
    [b,a] = butter(10, Wn, 'low'); 
% filter audio
    FXLChannel = filter(b, a, XLChannel);
    FXRChannel = filter(b, a, XRChannel);

% FUNCTION OUTPUT  
% Allocate left and right channel matrices.
    MTXleft = zeros(length(FXLChannel), 1);
    MTXright = zeros(length(FXRChannel), 1);
    MTXleft = FXLChannel;
    MTXright = FXRChannel;
    outputMTX = [ MTXleft , MTXright ];

wavwrite ( outputMTX, Fs, OUTfile );

end

[Geoff]

Oh, just cottoned on to your 'system overload' quote... Um, how large is this file? How much RAM and virtual RAM do you have?

[Geoff]

PS: When you post code in a question, please use the 'code' formatting button in the toolbar. Otherwise it comes out as a mess, my eyes glaze over, and I don't actually read it.

Answer 1

[Geoff]

I wouldn't be surprised if iTunes takes your computer down.

Have you tried plotting the waveform in MatLab? Does it look sensible, or does it perhaps look like it would challenge your hardware?

Have you tried loading inside a program that actually understands audio files, such as Adobe Audition? (Trial version available from their website - don't know if this is available for Mac). Or maybe open-source Audacity which I think is cross-platform.

A data file like this shouldn't kill your computer, even if the audio is complete garbage, unless perhaps:

(A) You have a serious issue with your sound hardware;

(B) You have a dead or pokey RAM module, and your WAV just happens to get loaded into it;

(C) The WAV file is too large to fit in RAM but your OS is too stupid to realise;

(D) You're opening the file with sucky software (eg iTunes)

(E) The file has been written into a corrupted area of disk that freaks your OS out;

(F) The use of this crosstalk algorithm is banned by alien police from the 5th dimension.

[James]

hey Geoff, thanks for taking a look. I can't plot the waveform in matlab. I get this error when I attempt a wavread of the file:

??? Error using ==> wavread at 166

Incorrect chunk size information in WAV file.

any thoughts based on that error message? I'm guessing my function isnt writing the file correctly, maybe there is a problem with the sampling rates being off? I will look into the other audio programs you suggested, however, I will say that I tried opening the file not in iTunes, just in my mac's Finder window, and I got the same issue. The file I was attempting to run through it is only 2.5 MB. The output file says its 1.7 MB, and the audio bit rate is, get this, -1003144256. ??? My guess is that letter F is whats going on here.

[Geoff]

Yeah, okay sounds like your file is corrupt. I've never used MatLab to create WAV files, but check the documentation for wavwrite. Looks like you're calling it incorrectly. The third parameter is supposed to be the number of bits per sample. You are passing the file name!! (That's supposed to be the 4th parameter). Also make sure your data is formatted correctly. I assume it's something like 16-bit stereo, which presumably means you need a 2-column matrix of uint16 values. Or something like that.

[James]

hmm... well the file I made is DEFINITELY corrupt. Even checking the file info in the finder window made my computer stop. The wavwrite doc says you can pass 3 parameters in and it will assume a bit depth of 16. Which I think is ok. But I will have to take a look at this later as I need to move on to another project Im working on right now. Thanks for your help so far Geoff, you're the best!!! If anyone knows how to get rid of corrupted files, let me know!!!

[Geoff]

Just delete it.

Answer 2

[Michael]

First thing I'd make sure of it that the variable you're writing to the WAV file is all real. Either take the abs() or real() of it.