
"I. R. Khan" wrote:
> 
> Hi all,
> 
> We are trying to make a hearing aid for profoundly deaf people. It is
> observed that if a sound signal modulated on an ultrasonic carrier is
> conducted through the bones in the human head, it is audible (and even
> understood in some cases) by profoundly deaf people. However, the exact
> phenomenon happening inside the head is not exactly known. We want to
> calculate the transfer function between the excitation position
> (position of the transmitter placed behind the ear) and the cochlea, to
> explain the phenomenon and improve the performance of our device. The
> problem is that we know the input but for the output (the sound actually
> heard), we have to depend on the human subjects and there are lot of
> differences in the data collected from different subjects. We find
> ourself stuck, and shall be extremely thankful if some one could please
> suggest a way to proceed.

It's pretty hard to guess what you might already know. 
	What I would think you would have explored is how individual frequencies
(both input and carrier) perform with individual subjects. At that level
the sound is either heard or it isn't (the subject responds or doesn't).
With enough data and at different frequencies for the same and different
subjects you should be able to come up with some mappings of what parts of
the process are linear and what are not.

The biggest difficulty I would thing is that like anything in your body if
you don't use it you lose it, so a great deal of the problem could simply
be the atrophy of neurological components. This I would guess lead to
significant changes in how subjects respond over time. It may take a
significant amount od stimulation over time before responses become
consistent and predictable. I would thinks subjects with recent hearing
loss of the type you are working with would be better candidates to start
with to remove that uncertainty.

-jim

----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----

Richard Owlett <rowlett@atlascomm.net> wrote in 
news:12p7jcag34a7b4e@news.supernews.com:

> Would an indirect measure be useful? I'm thinking a series of EEG's.
> 
> Carrier off, carrier on without modulation, modulated carrier.
> 
> Comparison to EEG's of normal hearing subject receiving modulation 
> signal thru normal hearing test setup.
> 

Yes, an evoked potential would be another viable approach.

-- 
Scott
Reverse name to reply

"Le Chaud Lapin" <jaibuduvin@gmail.com> wrote in 
news:1167292092.909584.288700@48g2000cwx.googlegroups.com:

> HRTF (Head-Related Transfer Function).


The HRTF does not take bone transmission into account-- just the shape of 
the ear and canal, the separation between the ears, and the acoustic 
shadowing of the head.


-- 
Scott
Reverse name to reply

"I. R. Khan" <ir_khan@REMOVE.hotmail.com> wrote in news:emvmvn$p8r$1
@aioe.org:

> Hi all,
> 
> We are trying to make a hearing aid for profoundly deaf people. It is 
> observed that if a sound signal modulated on an ultrasonic carrier is 
> conducted through the bones in the human head, it is audible (and even 
> understood in some cases) by profoundly deaf people. However, the exact 
> phenomenon happening inside the head is not exactly known. We want to 
> calculate the transfer function between the excitation position 
> (position of the transmitter placed behind the ear) and the cochlea, to 
> explain the phenomenon and improve the performance of our device. The 
> problem is that we know the input but for the output (the sound actually 
> heard), we have to depend on the human subjects and there are lot of 
> differences in the data collected from different subjects. We find 
> ourself stuck, and shall be extremely thankful if some one could please 
> suggest a way to proceed.
> 
> Regards,
> Ishtiaq.

You could take a psychophysical approach.

Alternatively, you could try to stimulate a DPOE (an otoacoustic emission) 
by stimulating at two frequencies through the bone and recording in the ear 
canal, looking for the third frequency.  

-- 
Scott
Reverse name to reply

I. R. Khan wrote:
> Hi all,
> 
> We are trying to make a hearing aid for profoundly deaf people. It is 
> observed that if a sound signal modulated on an ultrasonic carrier is 
> conducted through the bones in the human head, it is audible (and even 
> understood in some cases) by profoundly deaf people. However, the exact 
> phenomenon happening inside the head is not exactly known. We want to 
> calculate the transfer function between the excitation position 
> (position of the transmitter placed behind the ear) and the cochlea, to 
> explain the phenomenon and improve the performance of our device. The 
> problem is that we know the input but for the output (the sound actually 
> heard), we have to depend on the human subjects and there are lot of 
> differences in the data collected from different subjects. We find 
> ourself stuck, and shall be extremely thankful if some one could please 
> suggest a way to proceed.
> 
> Regards,
> Ishtiaq.


Would an indirect measure be useful? I'm thinking a series of EEG's.

Carrier off, carrier on without modulation, modulated carrier.

Comparison to EEG's of normal hearing subject receiving modulation 
signal thru normal hearing test setup.

I. R. Khan skrev:
> Hi all,
>
> We are trying to make a hearing aid for profoundly deaf people. It is
> observed that if a sound signal modulated on an ultrasonic carrier is
> conducted through the bones in the human head, it is audible (and even
> understood in some cases) by profoundly deaf people. However, the exact
> phenomenon happening inside the head is not exactly known. We want to
> calculate the transfer function between the excitation position
> (position of the transmitter placed behind the ear) and the cochlea, to
> explain the phenomenon and improve the performance of our device. The
> problem is that we know the input but for the output (the sound actually
> heard), we have to depend on the human subjects and there are lot of
> differences in the data collected from different subjects. We find
> ourself stuck, and shall be extremely thankful if some one could please
> suggest a way to proceed.

Your solution is, of course, limited to those people whose hearing
impairment is caused by a malfunction in the inner ear. The underlying
assumption seems to be that the cochlea works.

First of all, you need a model for how the cochle works, i.e. how the
movements of the filaments(?) inside the cochlea is transformed to
audible sound.

Second, you need a model for the anatomy of the ear and surrounding
areas. Whether you can get to one of those, is anybody's guess,
even if you can make measurements on some diseased person,
chances are that the different properties of living and dead tissue
will be important here. I wouldn't be surprised if the difference
between flowing and coagulated blood in the tissue, turned out to
be important.

Assuming you can get to a model of the anatomy, use a finite
difference or finitre element method to simulate how the
sound transmitted behind the ear affect the filaments in the
cochlea.

Depending on the wavelengths of the sound you transmit, be
prepared for multiple reflections inside the bone tissue of the
scull and/or the voids of the outer and middle ear.

Rune

Tim Wescott wrote:
>
> Wildly invasive would be to insert a microphone into someone's cochlea,
> and do your tests.

An approximation to this could be to fill a human skull with silly
putty or some such material of appropriate density (I assume the
density of the human head can be found :) ) in which a microphone array
is embedded in the anatomically appropriate place. How good would this
approximation be? I have no idea!

C

Thank you for your very informative and amusing responses. I think every
one is enjoying vacations.

A medical hospital is also involved in this research, but "the bosses"
want to keep surgery out of equation, at least for now. I will look into
your suggestions in detail, but right now I have been asked to study
FDTD method and find if it can be used for getting the transfer function
of head. Can any one please suggest some good book/paper about this
method? Any comments about using it in this problem will also be highly
appreciated.


Ishtiaq.

I. R. Khan wrote:
> Hi all,
>
> We are trying to make a hearing aid for profoundly deaf people. It is
> observed that if a sound signal modulated on an ultrasonic carrier is
> conducted through the bones in the human head, it is audible (and even
> understood in some cases) by profoundly deaf people. However, the exact
> phenomenon happening inside the head is not exactly known. We want to
> calculate the transfer function between the excitation position
> (position of the transmitter placed behind the ear) and the cochlea, to
> explain the phenomenon and improve the performance of our device. The
> problem is that we know the input but for the output (the sound actually
> heard), we have to depend on the human subjects and there are lot of
> differences in the data collected from different subjects. We find
> ourself stuck, and shall be extremely thankful if some one could please
> suggest a way to proceed.
>
> Regards,
> Ishtiaq.

Wack each subject in the head sharply with a wooden hammer, and take
the Laplace transforms of whatever results. ;)

Seriously,  I vaguely remember that people who work in speech
recognition face this problem often.  I think you are looking for
what's called an HRTF (Head-Related Transfer Function).
http://en.wikipedia.org/wiki/Head-related_transfer_function .  Seems
that people are always trying to come up with better models of the
head, although there seems to be as much interest in speech synthesis
as analysis. As I recall, modeling the head as H(s) is very hard
because:

1.  Head is non-linear.
2.  Head is time-variant.
3.  Transfer function varies widely from head-to-head.
4.  A true impulse would make the subject completely deaf. :)

(etc.)

So you have a non-linear system that is not only time-variant, but
there are many systems under test.

Nevertheless, if you want only one H(s) I would take a brute-force
approach.  Have the each subject swallow a microscopic tethered
microphone and generate white noise and measure the response.  Do this
for various situations, mouth open, closed, background noise, no
background noise (yes, very tedious).

Once you have collected enough data, find the optimum H(s) for
particular listening scenarios from the H(s)'s you measured. The
optimum H(s) could be the MMSE estimate.  You could also make your H(s)
listener-dependent.  Parameterized your H(s) and have the subject
subject tune the device by say, going to the WWW, playing a tone bank,
and putting their mouths over the speaker as the parameters are
determined. 

-Le Chaud Lapin-