Digital Signal Processing Reference
In-Depth Information
ity is frequency-dependent. Any components of audio signals whose level
is below the audibility threshold need not be transmitted; they are irrele-
vant for the human ear. Fig. 8.5. illustrates the general relationship of au-
dibility threshold versus frequency.
The next characteristic of the human ear that is of significance for audio
coding is a characteristic known as masking. For example, a sinusoidal
carrier at 1 kHz with constant amplitude is applied to the ear of a test per-
son, and the region around 1 kHz is investigated by applying other sinu-
soidal carriers, the frequency and amplitude of which is varied. It will be
found that the other test signals are not audible below a certain frequency-
dependent level threshold around 1 kHz. This is known as the masking
threshold (Fig. 8.6.). The shape of the masking threshold depends on the
frequency of the masking signal. The higher the frequency of the masking
signal, the wider the masked range.
Fig. 8.5.
Threshold of audibility
This characteristic of the ear is known as masking in the frequency do-
main (Fig. 8.6.). The relevant factor for audio coding is the fact that audio
components below a defined masking threshold need not be transmitted.
However, masking not only occurs in the frequency domain but also in
the time domain (Fig. 8.7.). A strong pulse in the time domain masks
sound signals before and after the pulse, provided the levels of these sig-
nals are below a certain threshold. This effect, and in particular premask-
ing, is difficult to imagine but very well explicable. It is due to the finite
Search WWH ::
Custom Search