help needed on dfilt.fftfir

I am pretty much failing to run the following code:

[song,f]=wavread('D:\song.wav'); song=(song(:,1)+song(:,2))/2; cut=song(100000:120000); s1=size(song,1); a=0; s2=size(cut,1); w=ones(s2,1); cc=xcorr(cut,song);

the song is ~12.9 million datapoints.

Errors:

??? Error using ==> fftfilt at 78 Filters of length greater than 2^20 are not supported. Use dfilt.fftfir instead.

Error in ==> xcorr>vectorXcorr at 127 pos_c = flipud(fftfilt(conj(y),flipud(x))); % positive lags

Error in ==> xcorr at 54 [c,M,N] = vectorXcorr(x,autoFlag,varargin{:});

Error in ==> code at 8 cc=xcorr(cut,song);

I searched for dfilt.fftfir, but could not properly use it. I tried using the bugfix on http://www.mathworks.com/support/bugreports/640328

This doesn't work because the attached file is possibly deleted. Could anyone check and confirm?

What should I do to correct this code? Kindly help.

Answer 1

[Wayne King]

Please see this bug report. It is the same error. There is a work around provided by MathWorks

http://www.mathworks.com/support/bugreports/640328

[Vishesh Vatsal]

Thanks. I ran this again. This filter-length error is thankfully not there now. Since the signal is relatively lengthy (~5 min), it seems the code cannot perform an fft now.

Error: ---------------------------------------------------------------------------------------------------------------------------------- ??? Error using ==> fft Out of memory. Type HELP MEMORY for your options.

Error in ==> fftfilt at 129 X = fft(x(istart:iend,:),nfft);

Error in ==> xcorr>vectorXcorr at 127 pos_c = flipud(fftfilt(conj(y),flipud(x))); % positive lags

Error in ==> xcorr at 54 [c,M,N] = vectorXcorr(x,autoFlag,varargin{:});

--------------------------------------------------------------------------------------------------------------------

Is there a workaround this problem or the memory problem cannot be fixed?