Ah sorry, that's on me for the misinterpretation.
I cannot hear them either but saying that, if you have really nice cleanly recorded (long samples) upright bass samples, try playing clustered harmony in the low range. You will hear them - 'sort of'. None of us can truly hear them clearly, I cannot either - the mud you are hearing is the harmonics becoming more audible is how it was explained to me. We are not hearing them in actuality but we are hearing them clash apparently. Asking me to hear much above that well - That's like asking me to hear 20Khz now and I am lucky to hear higher than 10Khz now.

 

La Monte Young's Well Tuned Piano is an excellent piece to get lost in overtones. I definitely can't identify them, but they're omnipresent.

 

There is some great stuff out there if we all look long enough and are patient enough. I probably cannot identify them either. We can recognise things when they clash or are sympatico though. That is where technology becomes extremely useful to tell us what it is we're (not) hearing I guess.

Also on the topic of interesting - There are two phenomenal young pianists I have been listening to, Aaron Parks and Tigran Hamasyan. Both their composition and improvisational skills are something else. They are like chalk and cheese and do both from a completely different place to each other but they are approaching music from what seems like a known place but it does not go where you might expect it. There's some stuff on youtube from both of them.

 

Thanks for the recommendations, I've been looking for piano improvisers other than Keith Jarrett and Paul Bley for a while now (not saying there aren't other improvisers I dig, but those two are my favorites).

 

Jarrett and Hancock for me and Bley is great too. Seeing you like Bley you should really enjoy Aaron Parks, he's like all of them rolled into one but with his own take on it. If you put "Aaron Parks Peaceful Warrior" with the inverted commas it should come up in youtube with others from the same CD on the right.

 

I think we've been suitably careful with words there. Hearing versus Identifying.
Subconsciously (at the level of our ears and perception), we do register all that overtone information. But it's just not useful to hear those overtones separately at a conscious level. Our brains have decided to use that overtone information to build timbre instead.
What's arriving at your ear starts out as a mashed up composite waveform. What is your brain supposed to do with that?
It's more useful to generate audio sensations where we say "I can hear and separately identify a flute and a piano (as distinctly different timbres) - rather than hearing a weird composite flutiano instrument playing those overtones as chords".
Software like Melodyne, and various stem extractors, etc , is still trying to replicate that separation trick which our brains subconsciously perform and we just take for granted.

 

with your ear it is easy, with graphics or number data it is quite hard.

it is not even sure that the highest peaks are the fundamental, and there are quite some acoustic instruments (or synthesizer sounds) which are difficult to measure for exactly that reason.

what the fundamental is, is only a matter of preception - and interpretation.

 

The pic in the OP is calculated from a single audio frame only.
Neither phase spectrum nor any information about audio before and after that frame is given so it's only helpful if the analyzed sound is a note or chord built from sinewaves or tones with very soft overtones - and even chords played on a Rhodes EP can be hard to detect.
With "real life sounds", you can add notes at certain frequencies with a certain level that won't be clearly audible but it'll be a different chord. That's why not even trained AI can safely detect that - too many different sounds and too many composers out there with too many ways to play a chord (notes at different velocities etc).
That's the beauty of music and sound.

 

i recently had some fun with a harpsichord piece in melodyne. it required a lot of preprocessing in regards of frequency filtering and dynamic processing before it worked at least halfway to detect everything correctly in polyphonic mode with melodyne.

i ended up using parallel tracks in another editor, one for detecting the transients, which is a bit delayed, and one for finding the frequencies, then superimposed the transients onto the frequencies. (like envelope following)

though the worst is when the attack (i.e. the note start) is not found and the rythm is incorrect.