Normalized Threshold Hearing Test

[CherylJosie]

I developed my own version of a hearing test that I'd like to share.

My version of hearing test doesn't give some difficult-to-interpret absolute threshold curve. I subtract out the Equal Loudness Contour threshold as graphed on Wikipedia and normalize the result to 0dB to obviate the need for an absolute reference level.

The expected result for 'normal' hearing is a straight line response at 0dB. Deviations in the measured frequency response threshold from the Equal Loudness Contour threshold appear as negative excursions from a straight line ('loss').

What it plots is 'what the world sounds like' before the brain adds its own learned 'perceptual correction' to mask hearing loss. Subtracting out the Equal Loudness Contour threshold gives the deviation from expected normal hearing and directly displays the filtering that is affecting the measured hearing sensitivity.

I'm hoping this hearing test gives a reference plot that multiple people can use to unambiguously compare the acuity of their hearing. It is also easy to administer. All it requires is a device to play the .flac files on, a receiver or other headphone amp with a calibrated volume control in dB, headphones, and a computer that can import a LibreOffice spreadsheet to plot the result. Yes, it's a manual process rather than an app.

I'll add replies in the top of this thread with attached files and a description of how to use them. I'll signify the end of my initial replies when I've completed the descriptions and uploads.

Until then, here's a plot of my own normalized threshold. I've had some osteoarthritic and vascular degeneration from aging, some ototoxic medication, and some SPL abuse over my lifetime, and my hearing shows it. Somehow, my hearing loss is symmetrical too and the curves for both ears are identical. I'm not sure how that happened and I didn't expect it, but that's what I measured. No, even though I used crappy ancient Tascam headphones, that's not the problem. It's definitely my ears. Okay well maybe the low frequencies aren't coming through the headphones well, but I've noticed some tinnitus in the low frequencies near 50Hz that I can turn on and off by clicking my teeth together, so I'm thinking it's a real dip.

 

[CherylJosie]

Here's the spreadsheet. In the spreadsheet I traced a copy of the Equal Loudness Contour with an XY plot to derive the level of the threshold in dB at the 31 band 1/3 octave center frequencies of a typical graphic equalizer.

 

[CherylJosie]

Here's the set of test tones. Each test tone has some tremolo to make it easier to distinguish from the constant drone of tinnitus.

I probably should have used vibrato instead. Oh well.

The tone files are too large to upload here so I'm linking to a YouTube video where the tones are concatenated into a sequential playback file.

I haven't personally used this video to test my own hearing. Based on what I heard when I tried playing it back on my home theater, I suspect that gain staging of this file may be problematic. Even with my volume control at maximum, I couldn't hear half of the tones in the video that I could readily hear through headphones.

I'm including the Linux script that I used to generate the tones, and the Linux script that I used to play them back on my PC. These scripts aren't required to run the test but they are provided for completeness and convenience. If you can't make the video work, maybe you can generate the tones yourself.

ffmpeg is available on Windows so people using that OS should also be able to adapt this script to generate the tones on their computers. Sorry, I don't use Windows so I can't create a script for you.

Oops, sorry, I forgot to include the link to the video. It's a playlist with one video generate from high res FLAC and the other from mp4. YouTube is going to mangle them both and they might not be useful. As of now the larger FLAC is still uploading. It should automatically post once it's done.

 

[CherylJosie]

So the procedure for using this test is as follows:

1) Play the video through headphones and adjust the gain on your playback device for maximum undistorted output on the first tone (20Hz).

With your balance set to full left or full right, set your dB-calibrated receiver volume control to a level where you can barely hear the amplitude-modulated tone pulsing on and off. Each tone plays for one minute to give you adequate time to set your level appropriately.

Record the level setting in dB in the appropriate column of the spreadsheet (B for the left ear, C for the right ear).

I recommend two passes, one for each ear. The full test takes about one hour to complete.

When you are finished entering your data, your normalized hearing loss frequency response is automatically graphed next to the data entry column. What you are seeing is the magnitude of relative filtering that your hearing exhibits at each frequency compared to a 'normal' threshold, minus the absolute value of attenuation in your threshold overall. Your maximum response is normalized to 0dB.

I found this plot to be much more informative and comprehensive than what my audiologist provided. Also, having the design of the hearing test and the amount of time/equipment dedicated to the testing process fully under my control provides superior results IMO.

Note that the dynamic range of this video exceeds 80dB. It's going to tax your system. The threshold of human hearing varies drastically with frequency. Be sure to set your gain staging for maximum undistorted 'preamp output' output on that 20Hz tone, or you won't have enough gain left on your master volume control of your headphone/power amp to check for dramatic attenuation in your treble. I also wouldn't recommend trying this test with directional loudspeakers because you might fry your tweeters and not even realize you are doing it.

 

[CherylJosie]

I'm done (for now) posting the hearing test and I'd like to open the post for discussion. There are a few things I'm most interested in.

One is whether or not anybody successfully uses my hearing test, and what their response looks like.

A second one is a theoretical interpretation of hearing loss. My ignorant understanding is that our perception automatically adjusts for hearing loss and we don't perceive it as it is actually manifesting. There seems to be some perceptual boost of modest hearing loss frequencies that preserves what we 'remember' as normal hearing so that masking is less problematic. I'm going to refer to this boosting as 'dynamic compression' because we still hear our threshold as an infinitesimal progression towards silence even though we are also hearing extremely loud sound as extremely loud despite our loss of perception.

It seems that only when there is profound loss will we perceive the threshold of pain as falling toward the threshold of hearing. Otherwise we seem to perceive nearly normal sound even with dramatic filtering.

I'm at a loss to explain how this happens.

One last item I'm curious about is distortion in damaged hearing. When I hear loud sounds in the midrange such as an emergency vehicle siren (ambulance in particular as they have a higher more piercing sound) I note a downward-sweeping artifact as the siren sweeps upward.

I have a friend who has gone the route of hearing aids who said his audiologist recommended he wear them to avoid his hearing becoming distorted. Apparently this audiologist believes that distortion arises primarily from psychoacoustic phenomena owing to loss of perception.

I don't see how that explains what I'm hearing as distortion. I'm thinking that what I'm perceiving is a frequency-shifted copy of the siren owing to nonlinear distortion from arthritic bones of my middle ear rattling against each other and resonating.

I doubt it has anything to do with aliasing. Our hearing doesn't digitize sound and there's no sampling frequency to cause aliasing. At least, I don't think so...?

Thank you for creating this forum and I look forward to interacting with various participants on my post. I haven't spent much time on here yet, but the technical discussions are first-class in Amir's technically rigorous tradition and I really enjoy that.