Compression of Audio Signals to MPEG and Dolby Digital - Digital Video and Audio Broadcasting Technology - page 144

The objective of audio compression is to reduce the 1.5 Mbit/s data rate
to between about 100 kbit/s and 400 kbit/s. MP3 audio files, which are
very widely used today, often have a data rate as low as 32 kbit/s. Simi-
larly as with video compression, this is achieved by way of redundancy re-
duction and irrelevance reduction. In redundancy reduction, superfluous
information is simply omitted; there is no loss of information. By contrast
in irrelevance reduction information is eliminated that cannot be perceived
at the receiving end, in this case the human ear. All audio compression
methods are based on a psychoacoustic model, i.e. they make use of the
"imperfection" of the human ear to remove irrelevant information from the
audio signal. The human ear is not capable of perceiving sound events in
the vicinity of strong sound pulses neither in the frequency nor in the time
domain. This means that, to the ear, certain sound events mask other sound
events of lower amplitude.
8.2 History of Audio Coding
In the year 1988, the MASCAM method was developed at the Institut für
Rundfunktechnik (IRT) in Munich in preparation for the digital audio
broadcasting (DAB) system. From MASCAM, the MUSICAM (masking
pattern universal subband integrated coding and multiplexing) method was
developed in 1989 in cooperation with CCETT, Philips and Matsushita.
MUSICAM-coded audio signals are used in DAB. MASCAM and
MUSICAM are both based on subband coding. The audio signal is split
into a large number of subbands, each of which is subjected to irrelevance
reduction to a greater or lesser degree.
At the same time as the subband coding method was developed, the
Fraunhofer Gesellschaft together with Thomson devised the ASPEC
(Adaptive Spectral Perceptual Entropy Coding) method, which is based on
transform coding. The audio signal is transformed from the time to the fre-
quency domain using DCT (Discrete Cosine Transform), and then irrele-
vant signal components are removed.
Both the subband-coding MUSICAM and the transform-coding ASPEC
method were included in the MPEG-1 audio compression method, which
was established in 1991 (ISO/IEC 11172-3 standard). MPEG-1 audio
comprises three possible layers: II essentially use MUSICAM coding, and
layer III principally uses ASPEC coding. MP3 audio files are coded to
MPEG-1 layer III. MP3 is often mistaken for MPEG-3. MPEG-3 was
originally aimed at implementing HDTV (high definition television), but
HDTV was already integrated in the MPEG-2 standard, so MPEG-3 was