Re: Speakers for speech testing (Christine Rankovic )

Subject: Re: Speakers for speech testing
From:    Christine Rankovic  <rankovic(at)ARTICULATION.COM>
Date:    Thu, 15 Sep 2005 10:14:20 -0400

Loudspeakers that perform within the tolerances given in Ward's note (below)
would not influence intelligibility substantially, according to the AI.

The hearing aid issues addressed by the van Dijkhuizen et al. studies are a
different matter altogether.  van Dijkhuizen et al. used frequency responses
that were considerably different from a flat response:  they were
intentionally shaped to both overcome hearing loss and suppress a
low-frequency band of noise.  For application to hearing aids, particularly
when there is background noise, frequency response slope and bumps can have
a significant impact on speech intelligibility, according to the AI, and
this is clear from speech recognition tests made by van Djikhuizen et al.
The AI's application to this problem was explored by Rankovic (1998)
"Factors governing speech reception benefits of adaptive linear filtering
for listeners with sensorineural hearing loss," Journal of the Acoustical
Society of America, vol 103, pp 1043-1057.  The impact of a frequency
response on speech intelligibility depends on precise values of the
parameters speech intensity, noise spectrum, and hearing loss.

Another AI finding is that is nearly impossible to "ensure audibility" when
there is any hearing impairment except, perhaps, in some quiet circumstances
for mild hearing losses.  This is due to the small residual dynamic range of
hearing-impaired listeners.

Christine M Rankovic, PhD.
Articulation Incorporated
AI = V x E x F x H


----- Original Message ----- 
From: "Brent Edwards" <brent(at)EDWARDS.NET>
To: <AUDITORY(at)LISTS.MCGILL.CA>
Sent: Tuesday, September 13, 2005 5:32 PM
Subject: Re: Speakers for speech testing


>From an AI perspective, the speaker response shouldn't matter as long as
audibility is ensured, the change in response doesn't affect spread of
masking relative to a flat response, and the rollover level isn't reached.
While not a directly answer to your question but related nonetheless, I had
this to say about the frequency response of hearing aids in my chapter of
the Springer speech book:

"The slope of the frequency response can change considerably and not affect
intelligibility as long as speech remains between the thresold of audibility
and discomfort (Lippman, et al. 1981; van Dijkhuizen et al. 1987), although
a negative slope may result in a deterioration of intelligibility due to
upward spread of masking (can Dijkhuizen et al. 1989)."

--Brent

----- Original Message -----
From: "Ward R. Drennan" <drennan(at)u.washington.edu>
To: AUDITORY(at)LISTS.MCGILL.CA
Subject: Speakers for speech testing
Date:         Tue, 13 Sep 2005 13:32:11 -0700

>
> Does anyone know of research that has investigated a effect of the
speaker's
> frequency response on speech perception ability? We could try to get
speaker
> response to be perfectly flat, but so long as we are within an ANSI
standard,
> does it really make a difference? Does anyone know the scientific basis of
> this standard?
>
> ANSI standard 3.6-1996 (from Katz on speech audiometry):
> No more than 10 dB attenuation 125-250 Hz
> +/- 3 dB 250-4000 Hz
> +/- 5 dB 4000-6000 Hz
>
> Ward R. Drennan, Ph. D.
> VM Bloedel Hearing Research Center
> University of Washington Box 357923
> Seattle, WA 98195-7923
> Phone: (206) 897-1848
> Fax: (206) 616-1828