Databases Reference
In-Depth Information
Bit Sliced Arithmetic Coding (BSAC)
In addition to the Huffman coding approach used in the MPEG-2 AAC scheme, the MPEG-4
AAC scheme also provides the option of using binary arithmetic coding. The binary arithmetic
coding is performed on the bitplanes of the magnitudes of the quantized MDCT coefficients.
By bitplane we mean the corresponding bit of each coefficient. Consider the sequence of 4-bit
coefficients
x
n
1. The most significant bitplane would consist of the MSBs
of these numbers, 011100. The next bitplane would be 100000. The next bitplane is 010110.
The least significant bitplane is 110011.
The coefficients are divided into
coding bands
of 32 coefficients each. One probability table
is used to encode each coding band. Because we are dealing with binary data, the probability
table is simply the number of zeros. If a coding band contains only zeros, this is indicated
to the decoder by selecting the probability table 0. The sign bits associated with the nonzero
coefficients are sent after the arithmetic code when the coefficient has a 1 for the first time.
Arithmetic coding is also used on the scalefactor information. The maximum scalefactor is
coded as an 8-bit integer. The differences between scalefactors are encoded using an arithmetic
code. The first scalefactor is encoded using the difference between it and the maximum
scalefactor.
:
5
,
11
,
8
,
10
,
3
,
17.5 Dolby AC-3 (Dolby Digital)
Unlike theMPEGalgorithms described in the previous section, theDolbyAC-3method became
a de facto standard. It was developed in response to the standardization activities of the
Grand
Alliance
, which was developing a standard for HDTV in the United States. However, even
before it was accepted as the recommendation for HDTV audio, Dolby AC-3 had already made
its debut in the movie industry. It was first released in a few theaters during the showing of
Star Trek IV
in 1991 and was formally released with the movie
Batman Returns
in 1992. It
was accepted by the
Grand Alliance
in October of 1993 and became an Advanced Television
Systems Committee (ATSC) standard in 1995. Dolby AC-3 had the multichannel capability
required by the movie industry along with the ability to downmix the channels to accommodate
the varying capabilities of different applications. The 5.1 channels include right, center, left,
left rear, and right rear, and a narrowband low-frequency effects channel (the 0.1 channel). The
scheme supports downmixing the 5.1 channels to 4, 3, 2, or 1 channel. It is now the standard
used for DVDs as well as for Direct Broadcast Satellites (DBS) and other applications.
A block diagram of the Dolby AC-3 algorithm is shown in Figure
17.14
. Much of the
Dolby AC-3 scheme is similar to what we have already described for the MPEG algorithms.
As in the MPEG schemes, the Dolby AC-3 algorithm uses the modified DCT (MDCT) with
50% overlap for frequency decomposition. As in the case of MPEG, there are two different
sizes of windows used. For the stationary portions of the audio a window of size 512 is used to
get 256 coefficients. A surge in the power of the high-frequency coefficients is used to indicate
the presence of a transient and the 512 window is replaced by two windows of size 256. The
one place where the Dolby AC-3 algorithm differs significantly from the algorithm described
for MPEG is in the bit allocation.
