The concept behind masking in audio is no big mystery: If you have two tones of different frequencies and levels playing simultaneously, the louder of the two will disguise the quieter one. That's a bit oversimplified, but when the loud tone is, say, ten times the volume of the soft one, it lets you get away with lots of things. Better still, the human ear is more sensitive to sound in a particular part of the audio spectrum, so the apparent loudness of tones in that region is higher than their actual volume. So what does this mean when we want to compress a digital audio file? Quite a bit (pun intended)....What MPEG compression does, in a nutshell, is divide the audio signal into 32 bands, or frequency regions. The masking effects of each band on the other 31 are calulated, and the encoder decides which bits can be tossed without much noticeable effect. The compression ratio can be adjusted, and there are weighting factors that take into account the response curves of the human ear. This "compare and save" process goes through the file and dynamically thins out the bitrate while preserving the most important information, but since a little bit of the baby goes out with the bathwater, we refer to it as a lossy scheme.
