i am doing a project on audio watermarking. in this after shaping the moise
spectrum according to the frequency mask of the human auditory system, i am
taking IFFT of the noise samples. the thing is that the number of frequency
bins for the unfiltered noise and the mask are both 512. so i will get 512 
points as the product of the above two . now i need to take ifft of 384
samples as per MPEG Audio Layer 1 model.But won't that distort and
compromise the information contained in the 512 points when only 384 point
IFFT is taken. what could be the solution to this?