Average Magnitude Differential Function - implementation

[John W. Eaton]

On  5-May-2010, jacek grabowski wrote:

| s=wavread('/home/jaca/ka.wav');
| w=s(2000:4000);
|  N=length(w);
|    for k=1:N
|        suma=0;
|        for n=1:N-k
|              suma=suma+abs(w(k)-w(k+n));    % 2 001 000 times for N=2001
|        endfor
|     amdf(k)=suma;
|   endfor
| 
| I wolud like to run it from PHP system('octave3.0 -qf amdf.m') but max
| execution time is only 30 s.
| When i run amdf.m direct on shell:
| 
| address@hidden:~$ octave3.0 -fq amdf.m
| Elapsed time is 58.2993 seconds.
| 
| It's any way to optimize this code?

Octave is an array language, so operations on arrays are generally
faster than explicit looping.  Your inner loop is fairly easy to
vectorize:

  amdf = zeros (1, N);
  for k = 1:N
    amdf(k) = sum (abs (w(k) - w(k+1:N)));
  endfor

Maybe someone else will have an idea for how to vectorize the
remaining loop.

jwe