On Thu, 28 Oct 2010, Ludwig Maes wrote:

Hi, I was wondering how one would locate a vowel in the IPA vowel quadrangle using pd... Of course I would expect the corners of the vowel to depend on individual characteristics, but I dont know where to start. Anyone any ideas? Greetings, Ludwig

I'd like to know as well... about two years ago, I made a patch to play Wikipedia's vowel samples according to three criteria (two [vradio] and one [tgl]) :

 0    1    2    3    4

6 [i][y] [ɨ][ʉ] [ɯ][u] 5 [ɪ][ʏ] [ʊ] 4 [e][ø] [o] 3 [ə] 2 [ɛ][œ] [ʌ][ɔ] 1 [æ] [ɐ] 0 [a][ɶ] [ɑ][ɒ]

And then some more. Personally I'm mostly interested in [iyɪʏeøeœaœəuʊøʌɑ] (the nonnasals used in my own speech), but I'd like a language-independent locator too.

Then for example I numbered the above like 001 = 6 0 1 = [ɶ] 200 = 4 0 0 = [ɛ] 511 = 1 1 1 = [ʏ] 531 = 5 3 1 = [ʊ]

The digits are : row number, column number, subcolumn number. I'd add a fourth number that'd toggle nasals (because I need [ɛ̃̃œ̃ø̃ɑ̃ɔ̃õ] to reproduce my speech) and perhaps a fifth number that would toggle [ː] but it might be more appropriate to write them as a pair of identical vowels (especially as some of them mutate to diphtongs in my speech).

I haven't looked much in the analysis part of it yet. I thought I could analyse the sound samples, extract their cepstrums, cross-fade them, resynthesise using a different envelope and tone... but I haven't tried any of that at all in the end (well, maybe I started trying it and it didn't work...).

| Mathieu Bouchard ------------------------------ Villeray, Montréal, QC

I had a go a number of years ago at coding an LPC-based formant detector for PD, based on the Speech Filing System from University College London. It has since remained on my "to do" list, since I got the object to load but there were problems in getting it working (i.e. it didn't, and crashed PD).

In theory, if you can detect the peaks in a Linear Predictive Coding-derive polynomial plot, then you can determine the vowel very accurately.

It would be nice to make this work - I'll dig it out and perhaps someone can help me understand why it doesn't work (and fix it). I'll dig it out and try to GDB it tonight, but I have to feed my child now.

Best, Ed Metastudio 4 for Pure Data - coming soon! Metastudio 3 still available at http://sharktracks.co.uk/puredata

----- Original Message ---- From: Mathieu Bouchard matju@artengine.ca To: Ludwig Maes ludwig.maes@gmail.com Cc: Pd List pd-list@iem.at Sent: Thu, 28 October, 2010 15:26:05 Subject: Re: [PD] IPA Vowel space analyzer

On Thu, 28 Oct 2010, Mathieu Bouchard wrote:

Then for example I numbered the above like 001 = 6 0 1 = [ɶ] 200 = 4 0 0 = [ɛ] 511 = 1 1 1 = [ʏ] 531 = 5 3 1 = [ʊ]

sorry, I screwed that one. I had gotten the first digit backwards, and I began to fix it and didn't finish. The correct list is :

001 = 0 0 1 = [ɶ] 200 = 2 0 0 = [ɛ] 511 = 5 1 1 = [ʏ] 531 = 5 3 1 = [ʊ]

else it wouldn't be consistent at all.

| Mathieu Bouchard ------------------------------ Villeray, Montréal, QC