[asterisk-biz] AQuA vs. PESQ

[sales at sevana.fi]

Dear List members,

We would like to share information about AQuA product we developed for
VoIP community that is used in Asterisk based voice quality monitoring
system we provide to perform end-to-end voice quality testing that does
not rely on traditional VoIP parameters like packet loss, jitter, latency.

Even now and then we get questions on how AQuA is similar / different to
PESQ. Well, AQuA surely differs from PESQ, because it’s using different
perceptual model, returns different parameters (AQuA returns prediction
for objective MOS score according to P.800 and PESQ values are from 1 to
4.5 considering that nobody can distinguish MOS 4.5 and MOS 5), has
different features, provides more information on reasons for audio quality
loss.

We clearly state that AQuA is not PESQ. There are cases which PESQ fails
to detect, but AQuA does catch the degradation, read this article
Perceptual Evaluation of Speech Quality (PESQ) (http://www dot microtronix
dot ca/pesq.html), and there are reports that state PESQ may not give
accurate scores for example in mobile and VoIP end-to-end testing (you can
google for “Limitations of PESQ”). However, the point is not that AQuA is
better than PESQ, it’s just different, but different in many ways.

We created AQuA to provide VoIP community with inexpensive tool to measure
and monitor voice quality in VoIP, mobile, PSTN and converged networks.
Currently we have customers that use AQuA for all types of networks
including sattelite communications and they are quite happy, because the
main point is that one needs to know whether the quality is good or bad
and if a tool delivers this information at a certain precision then it’s
useful.

When evaluating voice quality in VoIP or other communication network our
customers report two main issues:

1. test signal may have different level (RMS amplitude) than the original,
and in this case it requires amplitude adjustment that is done by free SOX
utility in AQuA 5.x. AQuA 6.x that is ready for release has built in
feature to adjust RMS amplitude using our in-house algorithm.

2. test signal may have “silence” in the beginning and / or end of the
recorded audio file. In order to synchronize two signals in time domain
one can use SOX as well, or use -trim parameter in AQuA, or even both.

There are a lot of parameters one can adjust in AQuA, but in most of the
cases they are set to default values and we deliver batch scripts for
Windows and Linux with pre-set parameters. Typical test then looks like

Windows: test.bat reference.wav test.wav
Linux: ./test.sh ./reference.wav ./test.wav

To summarize AQuA and PESQ do the same thing differently. If one requires
to provide its end customers voice quality score accordoing to
International standard, then it’s surely PESQ, because it’s a
recommendation from ITU-T P.862/P.863 bla bla bla that costs a lot of
money and is provided by different companies as a software / hardware
combination, boards, features, bla bla bla
 And if you are a VoIP service
provider or a mobile operator or DSP algorithms developer then AQuA is
your best choice – no royalties, no specific hardware (you want a dialer
connected to E1/T1, GSM or sattelite network? then use Asterisk VQM – open
source based voice quality monitoring soluton) – just pure voice quality
assessment.

If you have questions you are always welcome to contact us!

Best regards,
Sevana Oy

http://blog.sevana.fi/aqua-vs-pesq/

AQuA page:
http://www.sevana.fi/voice_quality_testing_measurement_analysis.php