NAL-NL2 wrapper, compression threshold

Post Reply
fdenk
Posts: 5
Joined: Wed Jan 11, 2023 8:38 am

NAL-NL2 wrapper, compression threshold

Post by fdenk » Tue May 14, 2024 9:16 am

Hi,

I am currently working on obtaining NAL-NL2 insertion gain targets using the nalnl2wrapper. First, it is great tool to have!
One input argument I was wondering about is the wide band compression threshold (wbct). If I understand the documentation of the original dll correctly, this value is usually computed based on user input regarding the hearing aid filterbank. However, it is passed to the wrapper as a dedicated input and set to a fixed 52 dB in the function gainrule_NAL_NL2.
Could you please explain the reason for this deviation from the original dll and the parameter choice made, and whether this might affect insertion gain outputs?

Thank you
Florian

tobiasherzke
Posts: 111
Joined: Mon Jun 24, 2019 12:51 pm

Re: NAL-NL2 wrapper, compression threshold

Post by tobiasherzke » Fri May 17, 2024 5:36 pm

Hi Florian, thanks for reaching out!

52dB was used for the wide band compression threshold because that value is the default in the NAL NL2 demo software "NAL-NL2 Clinician Edition". You can see this value in the tab "Target Screen" next to the label "Comp. Threshold".

That said,
  • I have not verified that this parameter in the Clinician Edition software is actually passed to the same wbct parameter of the CompressionThreshold_NL2 function.
  • I have not found any guidance in our NL2 developer documentation package on how to choose this parameter, other than 52 being the default value in the Clinician Edition. You seem to have found more information. I'd appreciate if you shared your findings, reasonings and/or conclusions in more detail, as I'm very interested in improving the nl2wrapper.
  • I have done some tests on some NL2 parameters to see if they have an influence on the proposed insertion gains. However, I do not remember if I have done those tests also for the wbct parameter. You could perform some tests in that direction if you like.
  • Users can change this parameter themselves in the Matlab/Octave m-file. If it makes sense, we can make it configurable through another mechanism, like a global Octave variable. But I do not see much benefit in this over editing the source code of the m-file. I'm still interested to change the default value in our distributed version of the m-file if another value makes more sense, or if another value can be computed from audiogram and hearing aid settings passed to the gainrule m-file!

fdenk
Posts: 5
Joined: Wed Jan 11, 2023 8:38 am

Re: NAL-NL2 wrapper, compression threshold

Post by fdenk » Wed May 22, 2024 1:41 pm

Hi Tobias,

thank you for the reply, meanwhile I have also noticed the 52 dB as default value in the NAL-NL2 clinical program. With this parameter and appropriate settings, we were able to exactly reproduce the output of the clinical program with the wrapper, so I would agree that the output is valid. I would also agree that there is no reason to make the wbct parameter easily changable by the user, I just wondered where the value comes from.

Upon re-reading the documentation (excel document called "NALNL2-Parameters", which we received with our package in the folder Documents), I realized that WBCT is actually passed to the dll as a fixed parameter and another parameter called "Compression Threshold_NL2" is computed from this. This actually makes much more sense.
Sorry for the resulting confusion!

Post Reply