DSP Robotics Support

I saw this little excerpt from a video (I've set the start time, please watch until approx. 16:00)
Does that mean that a vowel filter is basically just a bunch of bandpass filters with high resonance? Or does it need specific "sweetspots" that are hardcoded into a vowel filter?

https://youtu.be/qFpetPQ6usI?t=883

Yes you can get pretty good vowel filter from two bandpass filters, thats enough, I dont think you need more. There are some certain frequencies where they create a more recognizable human vowels, one has to experiment for the best spots

Ah, thanks a lot!

Vowel Formants is the search term you want - there are loads of linguistics sites with handy charts and tables giving you a starting point for frequencies to try out for typical male and female voices. None of them that I've ever seen show more than three bands - but as Adam says, the canonical diagram for the main vowels is usually only two-dimensional.

trogluddite wrote:Vowel Formants is the search term you want - there are loads of linguistics sites with handy charts and tables giving you a starting point for frequencies to try out for typical male and female voices. None of them that I've ever seen show more than three bands - but as Adam says, the canonical diagram for the main vowels is usually only two-dimensional.

Very useful tip! Thanks to you as well, good sir!

Hi ,
Say did you ever see my "Dual Peak Band pass filter"? Give it a try.....


viewtopic.php?f=3&t=36908&start=0&hilit=BobF

Later then, BobF.....

i am getting good results with the ircam chant library data, which supports 5 tessiturae, male and female, 7 vocals, 3 filters incl. bandwith coefficients.

i´ve used them with PAF, poles and zeros, and with resonator banks.

you start working with the data as is and then adopt the settings to your ears. but if you do that, make sure to play the tessiturae in the right keyboard range; if you play the bariton at 4000 Hz it wont work.