OPTICOM OPERA - V 3.5 User Manual page 147

C H A P T E R 6 : T E L E P H O N Y
T E S T I N G
Within OPERA the Waveforms of the signals before, as well as after the time
alignment can be shown. The signals are always directly derived from the raw
data and no filtering is applied. The waveform before the time alignment is the
raw signal as it is read from the WAVE file. The time aligned test signal, though,
is already processed by PESQ, which sometimes leads to unusual looking
situations at first glance. According to the P.862 standard, the time alignment
algorithm will repeat parts of the test signal during short dropouts. Although this
is very rare, the time alignment may also fail completely under some
circumstances, e.g. if you have a reference sequence consisting of the phrases a-
b-a and the test signal contains only b-a. In this case the time alignment
algorithm may mess up the phrases. When in doubt looking at both the aligned
as well as the unaligned waveforms will be helpful.
All analog equipment in particular introduces attenuation into the speech
signal. A high attenuation generally leads to a worse perception of speech.
PESQ does not weight this as a degradation, since it has no absolute reference
level available. Also, in real world systems, a low speech level on the electrical
side does not mean that the signal sounds very quiet, since the transducers
used, have a significant impact on the final loudness. For PESQ it is therefore
generally impossible to weigh the attenuation in terms of a perceived distortion.
The value of the attenuation however is important for optimizing the overall
system design. As delay, as well the attenuation/gain of modern telecom
equipment is not constant anymore, almost every mobile phone and VoIP
terminal has a built in AGC (Automatic Gain Control) or ALC (Automatic Level
Control). These mechanisms both target the same problem. They amplify or
attenuate the input signal before the transmission to compensate for very loud
or very quiet talker, and in this way keep the signal level in the optimum
operating range for the transmission. The gain adaptation happens constantly,
with a reasonable high time constant. Of course it is of major interest to not
only know the static behaviour of the transmission system, but also it's dynamic
characteristics. For this purpose our PESQ implementation provides both, the
overall attenuation in dB, as well as the gain/attenuation variation over time.
Both are derived from the time aligned and IRS filtered input signals. Out of
band energy below 300Hz is disregarded. The attenuation is calculated as the
ratio between the reference and the test signal energy. The gain variation is
updated only when one of the two signals exceeds the threshold in quiet.
Closely related to the attenuation is the measurement of the signal levels. Here
it is of special interest to know the signal levels separately for the active speech
parts of the signals as well as for the silent parts. Naturally it is important to
know these parameters for the reference as well as for the test signal, otherwise
it is not possible to see the influence of the device under test. In OPERA all
three parameters, the total level, the level of the active speech part and the level
of the silent parts (background noise) are shown for the reference as well as for
the test signal in dB . These parameters are derived from the time aligned and
ov
IRS filtered input signals. Out of band energy below 300Hz is not taken into
account. These values are exceptionally useful for the assessment of comfort
noise generators.
B A N D V O I C E Q U A L I T Y
141
Waveforms
Attenuation and
Gain
Level Measurement