Databases Reference
In-Depth Information
80
60
Masking
sound
40
20
-100
0
100
200
300
400
Time (msec)
F I GU R E 17 . 3
Change in the audibility threshold in time.
the audibility threshold for a brief interval preceding and following the sound. In Figure
17.3
we show the threshold of audibility close to a masking sound. Sounds that occur in an interval
around the masking sound (both after and before the masking tone) can be masked. The fact
that sounds that occur before the masking tone can get masked can be seen as evidence for the
inferential nature of human perception. If the masked sound occurs prior to the masking tone,
this is called premasking or backward masking, and if the sound being masked occurs after the
masking tone, this effect is called postmasking or forward masking. Forward masking remains
in effect for a much longer time interval than backward masking.
17.2.3 Psychoacoustic Model
These attributes of the ear are used by all algorithms that use a psychoacoustic model. There are
two models used in the MPEG audio coding algorithms. Although they differ in some details,
the general approach used in both cases is the same. The first step in the psychoacoustic model
is to obtain a spectral profile of the signal being encoded. The audio input is windowed and
transformed into the frequency domain using a filter bank or a frequency domain transform.
The sound pressure level (SPL) is calculated for each spectral band. If the algorithm uses a
subband approach, then the SPL for the band is computed from the SPL for each frequency
coefficient
X
k
. Because tonal and nontonal components have different effects on the masking
level, the next step is to determine the presence and location of these components. The
presence of any tonal components is determined by first looking for local maxima where a
local maximum is declared at location
k
if
2
2
and
2
2
. A local
|
X
k
|
>
|
X
k
−
1
|
|
X
k
|
|
X
k
+
1
|
maximum is determined to be a tonal component if
|
X
k
|
20 log
10
7
X
k
+
j
|
|
where the values
j
depend on the frequency. The identified tonal maskers are removed from
each critical band and the power of the remaining spectral lines in the band is summed to
obtain the nontonal masking level. Once all the maskers are identified, those with SPL below
the audibility threshold are removed. Furthermore, of those maskers that are very close to
each other in frequency, the lower-amplitude masker is removed. The effects of the remaining
