Digital Signal Processing Reference
In-Depth Information
However, acoustic-physical phenomena which have something to with the "inaccuracy"
of our hearing organs and the brain can be exploited:
In the case of audio-signals at least those parts of the signal can
be omitted in the encoder which are imperceptible to the human
ear as a result of its limited resolution capacity in the interplay of
the time and frequency domain and amplitude. This is referred to
as the
irrelevance reduction
of signals or data, i.e superfluous
information can be omitted without any loss of quality.
A compression process which achieves this must in consequence take into account the
psycho-acoustic properties of our hearing organs in the perception of audio-signals. As far
as the recognition of irrelevant information is concerned this includes:
•
The frequency response and the so-called threshold of audibility and
auditory sensation area (see Illustration 246) and
•
The masking effects which describe the inaccuracy of our hearing
organs (see Illustration 247).
The auditory sensation area in Illustration 246 shows the frequency-dependent perception
of amplitude. This also applies to the so-called threshold of audibility. According to this
our hearing is at its most sensitive around 4 kHz. This is determined by measuring the
point from what amplitude a frequency appears to be audible for a large number of
different frequencies.
Note: the volume L is a logarithmic measurement which is given in decibels (db).
If the amplitude uncreases by 20 decibels the amplitude has increased by the
factor 10 compared with the reference quantity, at 40 decibels by the factor 100
etc.
Usually the frequency scale is selected logarithmically in acoustics. Logarithmic calcula-
tion is exponential calculation. As a result the intervals on the frequency axis between
0.1 = 10
-1
and 1 = 10
0
, 10 = 10
1
, 100 = 10
2
, 1000 = 10
3
etc are equal.
Illustration 247 describes the so-called the masking effect. Imagine you are in a club.
Loud music blares from huge speakers. This is very hard work for the sense of hearing as
sound levels of 110 decibels and more are reached. Because of the extreme loudness of
the music it is almost impossible to talk unless you shout. In acoustics this is referred to
as masking. In order to remove this effect the sound level of speech has to be raised to
such an extent that it is no longer masked by the interference signal (in this case music).
In principle this describes the property of the ear which cannot hear weak tones in the
frequency context of a strong tone. This masking has greater bandwidth the louder the
tone in question.
