Compression of Audio Signals to MPEG and Dolby Digital - Digital Video and Audio Broadcasting Technology - page 155

In irrelevance reduction, subbands are also evaluated as to whether they
contain harmonics of signals belonging to a lower subband, i.e. whether
the masked signals are tonal (harmonic) or non-tonal components. Only
non-tonal, masked signals may be completely suppressed.
In MPEG coding, a certain number of samples are always combined into
frames. A layer I frame is formed with 12 samples for each subband. A
layer II frame is formed with 3 x 12 samples for each subband (Fig. 8.12.).
For each 12-sample block, the highest sample is determined. This sam-
ple is used as a scaling factor which is applied to all 12 samples of the
block to provide for redundancy reduction (Fig. 8.13.).
Highest value is used
for scale factor determination
for a block of samples
Block of
samples
Fig. 8.13. Redundancy reduction to MPEG-2 layer I, II
8.6 Transform Coding for MPEG Layer III and Dolby
Digital
Transform coding, in contrast to subband coding, uses no filter bank for
subband filtering; the splitting of audio information in the frequency do-
main is effected by a variation of the Discrete Fourier Transform. Using a
Discrete Cosine Transform (DFT) or Modified Discrete Cosine Transform
(MDFT), the audio signal is processed to give 256 or 512 spectral power
values. At the same time, in the same way as with subband coding, A Fast
Fourier Transform (FFT) is carried out with relatively high resolution in
the frequency domain. Controlled by the psychoacoustic model created