jonboley / simplecodec Goto Github PK

Suggestions from Alberto:

First of all you can simplify your code using only half of the FFT (you use the double spectrum for a long part of your code)
for example:
peak_count=2:length(centers(1:length(centers)/2))
instead of
peak_count=2:length(centers)
Also you have this line:
big_mask = max(big_mask, Schroeder(centers(peak_count)*(Fs/2)/N, fft_spl((peak_count)) , 14.5+bark(centers(peak_count)*(Fs/2)/N), N) );
which should have
fft_spl(centers(peak_count))
I think, funny enough it still works fine...
Also for some reason you divide by 2 at the end, which lowers the amplitude. It's not needed and you get normal amplitude
wavwrite(x3/2, Fs, decoded_filename);
becomes
wavwrite(x3, Fs, decoded_filename); %compensate for the lack of volume somewhere

Hmm, the masking threshold doesn't seem right.
I think it should be calculated as:

big_mask = max(A,Schroeder(Fs,N,centers(1)*Fs/N,fft_spl(centers(1)), 14.5+bark(centers(1)*Fs/N)));
for peak_count=2:sum(centers*Fs/N<=Fs/2)
    big_mask = max(big_mask,Schroeder(Fs,N,centers(peak_count)*Fs/N,fft_spl(centers(peak_count)), 14.5+bark(centers(peak_count)*Fs/N)));
end

but this seems to result in horrible audio quality...