Databases Reference
In-Depth Information
T A B L E 11 . 1
Performance of DPCM system
with different predictors and
quantizers.
Quantizer
Predictor Order
SNR (dB)
SPER (dB)
Four-level
None
2.43
0
1
3.37
2
.
65
2
8.35
5
.
9
3
8.74
6
.
1
Eight-level
None
3.65
0
1
3.87
2
.
74
2
9.81
6
.
37
3
10.16
6
.
71
the signal-to-prediction error ratio. These are defined as follows:
i
=
1
x
i
i
=
1
(
SNR
(
dB
)
=
10 log
10
(36)
2
x
i
−ˆ
x
i
)
i
=
1
x
i
i
=
1
(
SPER
(
dB
)
=
10 log
10
(37)
2
x
i
−
p
i
)
The results are tabulated in Table
11.1
. For comparison we have also included the results
when no prediction is used; that is, we directly quantize the input. Notice the large difference
between using a first-order predictor and a second-order predictor, and then the relativelyminor
increase when going from a second-order predictor to a third-order predictor. This is fairly
typical when using a fixed quantizer.
Finally, let's take a look at the reconstructed speech signal. The speech coded using
a third-order predictor and an eight-level quantizer is shown in Figure
11.9
. Although the
reconstructed sequence looks like the original, notice that there is significant distortion in
areas where the source output values are small. This is because in these regions the input to the
quantizer is close to zero. Because the quantizer does not have a zero output level, the output
of the quantizer flips between the two inner levels. If we listened to this signal, we would hear
a hissing sound in the reconstructed signal.
The speech signal used to generate this example is contained among the data sets accom-
panying this topic in the file
testm.raw
. The function
readau.c
can be used to read the
file. You are encouraged to reproduce the results in this example and listen to the resulting
reconstructions.
If we look at the speech sequence in Figure
11.7
, we can see that there are several distinct
segments of speech. Between sample number 700 and sample number 2000, the speech looks
periodic. Between sample number 2200 and sample number 3500, the speech is low amplitude
and noiselike. Given the distinctly different characteristics in these two regions, it would make
sense to use different approaches to encode these segments. Some approaches to dealing with
