Taking into account the direct path

Post Reply
tobiasherzke
Posts: 109
Joined: Mon Jun 24, 2019 12:51 pm

Taking into account the direct path

Post by tobiasherzke » Mon Oct 05, 2020 2:48 pm

This is the reproduction of a thread that was lost during a forum software upgrade failure. We have lost all forum posts posted between 01 Sep 2020 and 14 Sep 2020 during that upgrade. Below is a reproduction of the text of the lost posts:

Taking into account the direct path

Date: Tue, 01 Sep 2020 10:10:20 GMT
From: Mmar
I hope you do not mind that this may be a little bit off-topic.

We use the openMHA to apply gains prescribed by the DSL formula to wav files (offline processing). We then present the processed files to our subjects via headphones. However, compared to real hearing aids, this is missing the direct signal, i.e. the portion of the signal that is transmitted through the earpiece. This becomes especially important at low frequencies where the direct signal is often higher in level than the amplified signal from the hearing aid's receiver. This is because, at low frequencies, prescribed gains are often very low (sometimes at negative dB values) and attenuation of the earpiece for sounds from outside the ear (insertion loss) is close to zero. In other words, if we do not take into account the direct path, we leave out a significant portion of the low frequency content of the signal.

I wonder if someone of you has faced a similar problem. In this case, it would be great if you could share your ideas, thoughts and possible solutions. Or maybe someone could point me to a publication dealing with this topic.

Date: Fri, 04 Sep 2020 03:21:07 GMT
From: tobiasherzke
I understand that you want to simulate a system response to an input signal x

systemResponse(x) = compressor(x) + linearFilter(x)

and right now you have implemented

simulatedResponse(x) = compressor(x)

and are wondering how to add the missing

+ linearFilter(x)

There are many ways to achieve this. One example:

For duplicating the input signal channels have a look at openMHA plugin route.
For splitting the signal into different processing paths have look at openMHA plugin split.
For applying a linear filter have a look at openMHA plugin fftfilter.
For combining different signals have a look at plugin matrixmixer.

Date: Tue, 08 Sep 2020 14:28:26 GMT
From: Mmar
Thanks. Yes, you understand correctly. This is what we are planning on doing. Sorry, my first post was very vague.

Thanks for pointing me to the corresponding MHA plugins. This will become useful later in our project.

However, currently, we are still on a more conceptual stage.

One particular question we were thinking about was if simply adding the direct signal to the amplified signal results in levels that are unrealistically high (especially at low frequencies). After all, in our simulation we would be adding two signals that are perfectly in phase, which won't be the case in real hearing aids due to the signal processing delay imposed on the amplified signal. Of course, in the simulation, one could impose an artificial delay on the amplified signal, but this would probably introduce some nasty comb filtering effects when adding the two signals.

Moreover, with real haring aids, when the levels are increased by the direct signal, wouldn't that increased level be decreased manually by the hearing aid acoustician during the fitting process in order achieve the levels prescribed by the prescription formula? So maybe our entire simulation approach is wrong, since in reality the effects of the direct signal are compensated for anyways.

It would be great if someone could share their thoughts and ideas regarding this topic.
Thanks in advance.

Date: Thu, 10 Sep 2020 15:19:52 GMT
From: tobiasherzke
comb filtering effects
... are the same in the real world.

I think you are right that audiologists take into account the direct path, and I would think that they usually do so by selecting the correct ear canal fitting in the manufacturer's software. What the software does as a consequence to account for the direct path may be contained in their documentation or it may be a trade secret of the manufacturer.

Date: Mon, 14 Sep 2020 15:44:48 GMT
From: Mmar
... are the same in the real world.
Thanks for your reply. Yes, this is correct. You also have comb filtering effects when using real hearing aids, unless, maybe, your hearing aid is capable of some advanced signal processing to avoid (or at least reduce) them. But I do not think this is common (if it is a thing at all).
I think you are right that audiologists take into account the direct path, and I would think that they usually do so by selecting the correct ear canal fitting in the manufacturer's software. What the software does as a consequence to account for the direct path may be contained in their documentation or it may be a trade secret of the manufacturer
Yes, unfortunately, this often seems to be the case. The fitting softwares I used were not really transparent for the user when it came to the calculation of gains and corresponding correction values.

Post Reply