[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [AUDITORY] PESQ and other speech quality algorithms
I would to expose my own example since I use PESQ and other speech quality
algorithms for several years.
First, as Graham already mentioned, you can publish your PESQ estimations if
you use this algorithm for research purposes (not commercial). For instance,
I worked several years at Orange Labs where I used a licensed PESQ to
publish my own research. Then, I moved to a university where I used the free
version of PESQ.
Another important point; Stefan, are you sure that Loizou's PESQ
implementation fulfill the PESQ implementation guide (ITU-T Rec. P.862.3)?
If you did not check that (which takes time since you need the databases
included in the ITU-T Suppl. 23), you cannot really publish your result as
In any case, I think that it is not allowed to publish a Matlab version of
the PESQ algorithms. I would propose to any person interesting to use PESQ
to download the ITU C-code, to compile it and to verify is correct
Last, for any other person you are interested in free speech quality
algorithms, you can use the following ones:
- Frequency weighted segmental SNR
- Enhanced Modified Bark Spectral Distortion (EMBSD)
- Weighted Spectral Slope (WSS)
See an evaluation of these algorithms there:
N. Côté, V. Gautier-Turbin, and S. Möller, ?Evaluation of instrumental
quality measures for wideband-transmitted speech,? in Proc. 8th
ITGFachtagung Sprachkommunikation, Aachen, Germany, October 8?10 2008.
[Online]. Available : http://www.vde-verlag.de/data/prcd.php?docid=453120021
Institut d'Électronique de Microélectronique et de Nanotechnologie (IEMN)
Institut Supérieur de l'Électronique et du Numérique (ISEN)
41, boulevard Vauban - 59046 Lille
Tel : +33(0)3 59 57 44 26
Fax : +33(0)3 20 30 40 51
De : AUDITORY - Research in Auditory Perception
[mailto:AUDITORY@xxxxxxxxxxxxxxx] De la part de Paris Smaragdis
Envoyé : jeudi 25 juillet 2013 06:39
À : AUDITORY@xxxxxxxxxxxxxxx
Objet : Re: [AUDITORY] PESQ
As I understand it, the licenses you point to are either for
developing/using your own implementation of PESQ, or for using the official
implementation. If you use a third party implementation it is their
responsibility to obtain the proper licenses, not yours. Perhaps Graham, can
help clarify Opticom's position on this.
On Jul 24, 2013, at 5:44 AM, Stefan Bleeck <bleeck@xxxxxxxxx> wrote:
> Dear Omid, your doubt qualifies you as a scientist :) I could not
> believe it either...
> Here is the web site:
> These are the exact numbers (obtained from Opticom by email):
> PEXQ SW Suite incl. POLQA (executable) regular is 8.900,00 EUR ./.
> academic discount 30 %. (To clarify: it's a one-off, not per annum) It
> was made clear to me that I cannot publish results obtained using my
> PESQ code in an academic journal without a license.
> The algorithms are freely available (I got them from Loizou's
> fantastic "speech enhancement" book), but that apparently doesn't mean
> that we can use them for free...
> All of that now raises the question: what is the best speech quality
> algorithm that we can use for free. My current favourite: segSNR. But
> I am not sure any more that this actually IS free. What do you think?
> On 24 July 2013 05:44, Sadjadi, Omid <sadjadi@xxxxxxxxxxxx> wrote:
>> Can you point us to the document that has this information? I highly
>> The paper describing PESQ is publicly available (to IEEE members), which
means everyone can implement and use this metric:
>> A.W. Rix, J.G., Beerends, M.P. Hollier and A.P. Hekstra, "Perceptual
evaluation of speech quality (PESQ)-a new method for speech quality
assessment of telephone networks and codecs," in Proc. IEEE ICASSP, Salt
Lake City, UT, May 2001, pp.749-752.
>> From: AUDITORY - Research in Auditory Perception
[AUDITORY@xxxxxxxxxxxxxxx] on behalf of Stefan Bleeck [bleeck@xxxxxxxxx]
>> Sent: Tuesday, July 23, 2013 9:43 AM
>> To: AUDITORY@xxxxxxxxxxxxxxx
>> Subject: [AUDITORY] PESQ
>> I just found out that the _academic_ use of PESQ (perceptual
>> evaluation of speech quality, ITU recommendation P.862), that is
>> described with MATLAB source code in Loizou's Speech Enhancement
>> book, requires a license that costs around $7000 per year. I write
>> this as a warning for others against using it (unless you happen to
>> have the spare cash), because we didn't know this and did a lot of
>> work with it that we now can't publish.
>> Dr. Stefan Bleeck
>> Hearing and Balance Centre
>> Institute of Sound and Vibration Research
>> Faculty of Engineering and the Environment
>> University of Southampton, SO17 1BJ, UK
>> Room 4093, Tizard building (13)
>> bleeck@xxxxxxxxx Tel.: 02380 596682
> Dr. Stefan Bleeck
> Hearing and Balance Centre
> Institute of Sound and Vibration Research
> Faculty of Engineering and the Environment
> University of Southampton, SO17 1BJ, UK
> Room 4093, Tizard building (13)
> bleeck@xxxxxxxxx Tel.: 02380 596682