This part of the encoder (sometimes referred to
as a Psychoacoustical Model or the Masking Model) analyzes the incoming
signal and outputs values
based on well-known pyschoacoustical phenomena. This is a critical
stage and has a large play in the overall quality of the encoder.
It has two parts. First it processes the signal
with a 1024 point FFT
.
Second, the resulting information (phase and amplitude) is analyzed.
From this analysis, the Perceptual Model then accomplishes
its two main functions. First, it tells the MDCT
whether
to use long or short windows. Essentially, it looks for transients,
sharp frequency
spikes that often occur at the onset of a sound. Transients equal
shorter windows, steadier sounds equal long windows. Second, it looks
for frequencies that might be masked
and outputs a masking threshold to the quantizer.