The game Rocksmith and some polyphonic guitar tuners (like the PolyTune) are able to recognize, at least to some extent, multiple fundamental frequencies in single-source audio.

I know there are some methods for multiple fundamental frequency detection, like MUSIC and ESPRIT, but is it likely that these products implement them or leverage several assumptions (such as: audio is from a guitar, audio is from a single source) about the application domain to make a simpler method work?

In the second case, which simpler methods are likely to be used? Simple peak-picking on a spectrum, with domain-specific heuristics? Or something more complex?

  • 1
    $\begingroup$ i've always been suspect of the efficacy of the multitone pitch detectors. i think lotsa assumptions are made. open strings? significant energy in the fundamental? i still dunno how it keeps the 4th harmonic of the E6 string from messing with the fundamental of the E1 string. try tuning those together. $\endgroup$ – robert bristow-johnson Jan 19 '15 at 17:10
  • $\begingroup$ The tuner may be able to assume open strings, but Rocksmith can't. I'll try to tune E1 and E6 together and see if the tuner can tell them apart. $\endgroup$ – Rafael Almeida Jan 19 '15 at 19:38
  • $\begingroup$ @robertbristow-johnson I ended up forgetting to post here, but I tried tuning the 6th and 1st string together and it worked well. $\endgroup$ – Rafael Almeida Mar 15 '15 at 20:08
  • $\begingroup$ dunno what Rocksmith does. most monophonic pitch detection algorithms will run in the time domain (i.e. no FFT). they might have sharp (high Q) BPFs tuned to the correct fundamental of each string. or perhaps two BPFs per string, one slightly sharp and the other slightly flat. and, perhaps some autocorrelation going on to give the algorithm some idea of what the pitch is when it is far from correctly tuned. $\endgroup$ – robert bristow-johnson Mar 15 '15 at 20:29

Check out chapter 1.3 of this IRCAM paper on multi-F0 estimation. It discusses the difficulties in extracting multiple F0s from a recording, including the handling of overlapping partials, transients, and reverberation, as well as the modeling of domain-specific sources with varied spectral properties.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.