[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Phonetic balance

Dear Auditory List,

There is some work by a student of mine, along with a follow up study, that may lead you to the conclusion that "phonetically balanced" mono-syllables may be the wrong way to measure speech perception, especially in the hearing impaired ears.

What this research shows is that each spoken token has a threshold, which we call SNR90 (and some times, SNR50). This number is the SNR in dB, at which the token score goes from 100% correct (zero error) to 90 % correct (10% error, or 1 error in 10 trials).

It turns out, and this seems a surprising result, that if you take each token, and normalize the error probability of error vs SNR function [Pe(SNR) = 1- Pcorrect(SNR)] by the SNR50 value, namely Pe(SNR-SNR50), then you get what looks like a step function for the ensemble average of all the tokens.

The distribution of SNR50 values has a large variance. The slopes of these normalized error curves have a very steep slope, typically (10%/dB). The ensemble average of all these step functions gives you the articulation index formula for the error, exactly. Namely, this analysis explains the articulation index.

Once you realize what is going on, it changes the entire way you look at the problem.

My conclusion is that "phonetically balanced" is the wrong way to quantify speech perception, as it is working with the wrong variable (phone vs SNR50).

Look at the data and decide for yourself. I will not try to convince you that this view is right, or wrong.

1) Riya Singh and Jont Allen (2012); "The influence of stop consonants’ perceptual features on the Articulation Index model," J. Acoust. Soc. Am., apr v131,3051-3068 (pdf <http://hear.ai.uiuc.edu/public/SinghAllen12.pdf>)

2) Toscano, Joseph and Allen, Jont B (2014) /"Across and within consonant errors for isolated syllables in noise,/" Journal of Speech, Language, and Hearing Research, Accepted July 25, 2014; doi:10.1044/2014_JSLHR-H-13-0244, (JSLHR <http://jslhr.pubs.asha.org/Article.aspx?articleid=1894924>)

These references are available at http://hear.beckman.illinois.edu/Main/Publications

Jont Allen
Univ of IL, Urbana IL

On 11/14/2014 02:21 AM, Brian FG Katz wrote:

Here is a work we did some years ago, creating a dataset of phonetically balances French. The methodoly description could be of some help to you.

A. Raake and B. Katz, “SUS-based method for speech reception threshold measurement in French,” in/Fifth Conf. on Language Resources and//Evaluation (LREC)/, (Genoa, Italy), pp. 2028–2033, May 2006,(url) <http://www-sop.inria.fr/reves/OPERA/publis/LREC06.pdf>.

Alexandre Raake did the real work, so please contact him if you have any questions.

Good  luck,



Brian FG Katz, Ph.D, HDR

Audio & Acoustique

Rue John von Neumann
Campus Universitaire d'Orsay
Bât 508
91405 Orsay cedex


Phone. +  33 (0)1 69 85 81 55 - Fax.  +  33 (0)1 69 85 80 88
http://www.limsi.fr <http://www.limsi.fr/> web_group: http://www.limsi.fr/Scientifique/aa/ web_theme: http://www.limsi.fr/Scientifique/aa/thmsonesp/

Ce message et toutes les pièces jointes sont confidentiels et établis à l'intention exclusive de ses destinataires.
Toute utilisation ou diffusion non autorisée est interdite.
This message and any attachments are confidential and intended solely for the addressees.
Any unauthorised use or dissemination is prohibited.

*De :*AUDITORY - Research in Auditory Perception [mailto:AUDITORY@xxxxxxxxxxxxxxx] *De la part de* Michal Soloducha
*Envoyé :* mercredi 12 novembre 2014 14:52
*À :* AUDITORY@xxxxxxxxxxxxxxx
*Objet :* Phonetic balance

Dear colleagues,

I search for some software which would help me to make my speech material phonematically balanced. Is anybody aware of any solutions or libraries (e.g. Matlab) which could be used for that?

Best regards,
Michał Sołoducha


Michał Sołoducha, Ph.D. Student

Assessment of IP-based Applications

Telekom Innovation Laboratories (T-Labs)

Technische Universität Berlin

Ernst-Reuter-Platz 7

10587 Berlin, Germany

Phone: +49-30-835358384

Fax: +49-30-835358409

Email:michal.soloducha@xxxxxxxxxx <mailto:michal.soloducha@xxxxxxxxxx>

url:www.aipa.tu-berlin.de <http://www.aipa.tu-berlin.de/>