Sound: To Formant (robust)...

A command that creates a Formant object from every selected Sound.

Settings

The settings for Time step (s), Maximum number of formants, Maximum formant (Hz), Window length (s) and Pre emphasis from (Hz) are as in Sound: To Formant (burg).... The following settings determine aspects of the iterative formant frequency refinement.

Number of std. dev.,
determines the number of standard deviation from where selective weighing of samples starts.
Maximum number of iterations,
determines the maximum number of iterations allowed in the refinement step.
Tolerance,
detemines another stop ctriterion for the refinement step. If the relative change in variance between successive iterations is less then this value, iteration stops. Iteration stops whenever one of the two defined stop criteria is reached.

Algorithm

First the sound is downsampled to twice the maximum formant frequency. Next the LPC coefficients are determined by the autocorrelation method. Finally, in an iterative procedure as described by Lee (1988) the formant frequencies and bandwidths are refined by selectively weighting of samples values.


© djmw, October 27, 2011