I've refined procedures in pitch tracker [snac~] (based on Philip McLeod's Specially Normalized Autocorrelation function (SNAC)), and renamed the object in the honour of the great 19th century explorer in the world of sound and frequencies, Hermann Helmholtz. The guy who listened through resonators to isolate frequencies in a sound.

[helmholtz~] is by default deaf for noise, but hears periodic sound no matter how faint. As a result, analysis frames containing transients, note jumps, or other signal segments with low autocorrelation are ignored. It's better to have no pitch report than an erroneous pitch report. Pitch estimation is precise up to +/- 0.1 Hz.

While regular methods need at least two periods in the analysis frame for detection, [helmholtz~] can do with a minimum of 1.2 periods. As a result, a female voice can be pitch-tracked with a frame of 512 samples at SR 44100, that is ~11 milliseconds. For a male voice and most instruments, 1024 samples are sufficient, that is ~23 milliseconds.

Remember that a time domain method can not distinguish multiple pitches in a stream, so [helmholtz~] must be used with monophonic channels only.

The object uses 0.6% CPU time per instance on my 2GHz CPU, with default settings, no analysis overlap. Since it only needs fundamental frequencies, it can be used in downsampled time as well, for example with resampling factor 0.25. Then you can do four overlapping analyses for the price of one, making things smoother as you get analysis results at a faster rate. Don't forget to resize the analysis framesize accordingly, otherwise you get long latency times.

I've extensively tested [helmholtz~] with my voice and all sorts of sweeping and jumping test tones. It does not seem to make mistakes with these sources. That may be different for other instruments, I'm eager to hear comments.

Attached file helmholtz~.zip contains source code, makefile, 32 bit binaries for OSX, Windows and Linux, help file and test patches.

Thanks to elden, who posted so many useful links on pitch shifting and tracking on this forum. [helmholtz~] would not have existed without all the info's. Find Philip McLeod's thesis introducing the SNAC function here:

http://miracle.otago.ac.nz/postgrads/tartini/papers/Philip_McLeod_PhD.pdf

Katja

P.s. [snac~], the experimental predecessor of [helmholtz~], is now discontinued.

Edit: find the latest [helmholtz~] now on:

http://www.katjaas.nl/helmholtz/helmholtz.html

http://www.pdpatchrepo.info/hurleur/helmholtz~.zip

Hello diemildefreude,

If you set [helmholtz~] to give output at lower fidelity level, for example [fidelity 0.5(, you will see that fidelity of the second note is much lower than [helmholtz~] default fidelity setting of 0.95. Also, you will notice that the pitch is actually detected with that lower fidelity setting. If you use patch [helmholtz~-soundfiletest] which comes with the helmholtz~ package, you can see the autocorrelation plot for the notes, and it becomes clear that the second note is particularly 'distorted', has lower harmonic content. This sound is typical for trombone, but you could experiment with fidelity settings to find the appropriate setting for trombone.

Katja

Hey guys, sorry to ask a question that I could just test myself, but my old MacBook recently died and my new laptop isn't compatible with my audio interface. But I'm curious if the beat frequency can be detected from a single note on a guitar string. Because the vertical restrictions of vibration at the bridge and nut of a guitar are different than the horizontal ones, the transverse waves along a string can travel at different speeds depending on their angle. This causes slow and subtle beating. I was trying to approximate this a while ago by using two detuned waveguides played together, but I was just sort of playing it by ear since I didn't know how to determine the beat frequency (and I have a feeling it may be amplitude-dependent and thus changing as the note decays; there is a frequency drop as the note decays as well)

Anyway, I'd love to hear if you all think [helmholtz~] can do this. Ricky, I know you've been working on analyzing features of guitars for your live performance setup. Maybe you could take a look? Like I said, I'd try it myself, but I need to save up for a new interface. If you can, thanks in advance.

I will be revisiting this at the end of June. I don't see why it wouldn't be possible? I will certainly look into it.

Thanks for your reply. It works even with [fidelity 0.8( , though I think I will put in 0.7 just in case.

Hi,

i tried [helmholtz~] and it works great but i need pitch detection for high frequencies which is greater than 17000. I couldn't update [helmholtz~] for detecting high frequencies. How can i disable lo pass filter and detect high frequencies with [helmholtz~]

@ryback3 said:

i need pitch detection for high frequencies which is greater than 17000

Sorry, the technique used in [helmholtz~] (autocorrelation) is very precise but not for high frequencies. Certainly not 17 KHz. Pitch of acoustic instruments rarely go above 1 KHz, that's [helmholtz~] default limit but it can go to 2 KHz using message maxfreq. If you want higher, better use an fft based detector like [sigmund~] or [fiddle~].

Katja

@katjav said:

@ryback3 said:

i need pitch detection for high frequencies which is greater than 17000

Sorry, the technique used in [helmholtz~] (autocorrelation) is very precise but not for high frequencies. Certainly not 17 KHz. Pitch of acoustic instruments rarely go above 1 KHz, that's [helmholtz~] default limit but it can go to 2 KHz using message maxfreq. If you want higher, better use an fft based detector like [sigmund~] or [fiddle~].

Katja

I tried sigmund~ and set -maxfreq to 22000 and npts to 1024 but i couldn't get pitch value which are greater than 15000 hz, what is the problem.

i couldn't get pitch value which are greater than 15000 hz, what is the problem

Not sure what use you have for sounds with pitch so high they are inaudible for most humans, but anyway you could try [sigmund~] with frequency peaks instead of pitch (see it's help file for details). Note that a 15 KHz sound can not have harmonics with standard samplerate 44.1 KHz, or they will be aliased. So you could just as well track the frequency component with highest amplitude.

Katja

On a monophonic traditional song audio file (female voice), I tried fiddle~, dfreq~, sigmund~, helmhotz~. With default parameters, the best result I got for far is with [helmhotz~]. Thank you Katja !

Hey everyone...I just now getting some time to play with this, and it is indeed pretty interesting in comparison to sigmund~ and fiddle~...I run into the same problems though, in trying to figure out how to filter data...ricky mentioned using [round] and I thought to try that, but I don't seem to have that object, what library is that in?

Hey!

[expr rint($f1)]

That'll do it!

Ricky

Hello Katja and/or whoever

Sorry for such a low-level question, but I'm not very good with these things.

I downloaded Helmholtz~ and loved it! It worked immediately on my Linux (Crunchbang) laptop. My main audio computer, though (by far, the best I have - and the one I'm using for my current project) is a Mac G5 with a PowerPC processor (got it used for $100!)

And I can't get helmholtz~ to work on it.

Is this because it's made for Intel processors? If so, could I make a ppc version using the LibraryTemplate I just downloaded, on the working Linux system? I've spent a few hours trying to answer this myself.

If it's not possible, that's good to know. I can never tell the difference between my incompetence, and computers not being the magic boxes of fairy tales.

Chris

Looking at the makefile with the [helmholtz~] package, I see that it doesn't do a 'fat binary' for OSX, but a native build for the platform (which was Intel Mac 32 bit in my case).

I suspect that the library template makefile does not facilitate cross-compiling. It would be much easier to build on the G5 using the included makefile. However, you may not have developer tools installed on that computer?

Katja

I'm afraid I'm in over my head here. I still spend half my time connecting signal outlets to non-signal inlets.

However, I'm happy to give it a try, if I'm trying the right thing. I don't yet know how to "build on the G5 using the included makefile," but I'll try to download and learn. May just end up using sigmund~ for this project.

It certainly seems like a question for another forum now... you guys are all here talking about cool stuff like auditory perception and trombone tracking, and all I can offer is "My cat's breath smells like cat food."

@Load074 said:

I don't yet know how to "build on the G5 using the included makefile,"

What do you mean exactly? If you have Apple's developer tools installed on your G5, it should be simple. In the [helmholtz~] project from http://www.katjaas.nl/helmholtz/helmholtz.html, just follow the section 'how to build' in README.txt. Basically it is a matter of typing 'make' on the command line.

However, if you don't know how to compile at all on the G5, you'd have a steep curve to follow. It may even take a while to find developer tools if you don't have them. Still I think it's worth the effort to try, not so much because of [helmholtz~] alone, but because you may find more open source software that you want to compile for it. On recent OSX versions it is no longer possible to compile for ppc, so you may find yourself on your own more often. There happened to be a Pd-list discussion on ppc support not long ago:

http://lists.puredata.info/pipermail/pd-list/2013-10/104676.html

By the way which processor type is in your G5? I could try to produce a fat binary of [helmholtz~].

You're on to me. It's true, I don't know, at all, how to compile. And I do plan to learn (or get an up-to-date computer), but I haven't yet. As you said, steep curve... totally worth it... maybe more than I'm able to bite off right now. Self-education via Google is arduous.

My processor is Dual 2GHz PowerPC G5, running 10.5.8. Any assistance is appreciated, but certainly not expected.

Hi @katjav! Can the helmholtz~ object output the sinusoidal peaks (frequency and amplitude) of an incoming audio signal?