MPEG Audio is a family of generic standards for low bit-rate coding
Layer I
low complexity, good for consumer recording
Layer II
high efficiency with medium complexity, good for professional recording and for broadcast
Layer III
high complexity and high efficiency, suitable for very low bit-rates application

key aspects
Input and output signals
AES/EBU
sampling frequencies: 32, 44.1 and 48 kHz
sample representation: PCM max. 24 bits
Coding modes
mono
stereo
joint stereo
dual channel
Bit-rates
from 32 Kbps to:
448 kbps for Layer I
384 kbps for Layer II
320 kbps for Layer III
Coding scheme
masking threshold computation (based on the perceptive model) on a 1152 samples window (384 for Layer I)
signal subdivision in 32 subbands
subband bit allocation according to the masking threshold
Packet structure
Fixed length

MPEG-1: high quality digital audio coding
MPEG1 audio provides:
variable compression factor
different complessity layers
compact disc quality at compression factor ~6
possible use in many applications (broadcasting, telecom, recording, multi-media, etc..)
Compact Disc Quality
The reproduced signal quality must be subjectively indistinguishable from the quality obtained with a 16 bits PCM system sampled at 44.1 kHz (compact disc) in the majority of normal use programs.

MPEG-2: the multi-channel extension
MPEG2 audio extension allows:
compact disc quality
environment sonority reconstruction
sonorous sources location
multilingual transmission
ancillary services
additional freq. 16000Hz, 22050Hz and 24000Hz
MPEG-Multichannel does not only support 5.1 audio, but also 7.1 audio

MPEG-2.5 Audio
additional frequencies added: 8000Hz, 11025Hz and 12000Hz



Layer II multi-channel characteristics
dynamic crosstalk
MPEG-Multichannel does not only support 5.1 audio, but also 7.1 audio

sounds that don't contribute to the source location can be transmitted by any channel

adaptive prediction

an adaptive predictor is used to reduce the inter-channel redundancy
the multi-channel extension channels are predicted from the two base channels
the prediction coefficients can be computed and transmitted at each frame

dynamic channel switching

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

the perceptive coding
The studies about the human ear sensitivity to the different frequencies and the different masking effects yielded a perceptive model on which the audio perceptive coding is based.


The static masking
The minimum perceived sonorous pressure, called threshold pressure, depends on several factors, as the source direction, the presence of other sounds, the kind of sound and especially the frequency. To exploit the static masking the threshold pressure must be known for each frequency. It must be noticed that the threshold is subjective and has a statistical value.

The dynamic masking
The dynamic masking is the effect of the threshold pressure increase because of the presence of another sound.
Simultaneous masking
When the masking sound and the masked one occur at the same time. The masking effect depends on the intensity and frequency of both sounds.
Non simultaneous masking
It's possible that the masking sound occurs before (sometimes after) the masked one. Actually the masking effect can last more than the sound by which it's caused.

The noise allocation
Because of all the masking effects the human ear is able to perceive only a part of the sound spectrum. The perceptive model allows the computation of the masking thresold for a defined samples set.
In the perceptive coding the bit-allocation is performed looking at the Signal to Mask Ratio (SMR) obtainable. In such a way bits aren't wasted representing sounds that wouln't be perceived.




Acknowledgments
The help of Giorgio Dimino was essential for providing the information.