Display modal pitch in details

[alk224]

As per Brown et al (2000), modal frequency is more important than mean frequency in the perception of "femaleness" according to voice. While average is no doubt a useful metric to inform the user of your app (which I love and use regularly), it would be very helpful if the recording details also included modal frequency (the most commonly repeated frequency during speech).

Brown M, Perry A, Cheesman A D, Pring T. 2000. Pitch change in male-to-female transsexuals: has phonosurgery a role to play? International Journal of Language & Communication Disorders 35: 129–136.

[mccartykim]

So would a loose algorithm for measuring this be to divide the voice recording into samples, measure each sample's frequency, and then declare the modal frequency whichever range of frequencies was most common? I'm having trouble finding a concise description of it in either the paper you cited or a Google search, but if it's that simple, I'm curious if I could implement it.

[alk224]

I think that would work well. If I were to guess, speech language pathologists determining modal frequency likely use commercially obtained software for their recording analyses, which probably implement determination of modal frequency in a variety of ways. As such, description of modal frequency is unlikely to be discussed in a paper like the one I cited, and I'm not terribly surprised Google isn't helping either. Your suggestion seems like it would probably work fairly easily as long as the sample divisions were fairly short, but I think you probably know better than I. The simple answer as far as I know is to bin all of the frequencies observed across a recording, and report the one most often observed. I guess this leads into your suggestion since said bins would require some decision on bin size (duration).

[mccartykim]

I made a quick and dirty fork of somebody else's JS pitch tuner to try to get the general idea. Not sure if my bins are too big. Also, both the original repo and mine don't seem to work on Firefox Quantum, which was disappointing to discover after I got to work.

https://mccartykim.github.io/voice-analyzer/index.html

[alk224]

I'll be able to try this out later today. Thank you so much for pursuing this! I'm trying to get some references together about modal versus mean pitch for my SPL and can pass the citation list to you if you want to include it anywhere in your documentation.