cVSA: Compute the Vowel space area using continuously measured...

In FredrikKarlssonSpeech/articulated: This package adds

Compute the Vowel space area using continuously measured formant frequency values

Description

This function implements the algorithm proposed by \insertCiteSandoval.2013.10.1121/1.4826150articulated, which takes continous measurements of formant frequencies for a speaker and computes the vowel space area from the convex hull of them, once spurious measurements have been removed using a likelihood threshold applied to the result of a Gaussian mixture model (GMM).

Usage

cVSA(
  F2,
  F1,
  vowel_categories = 5,
  threshold = 0.3,
  center = FALSE,
  scale = FALSE
)

Arguments

F2	A vector of F2 formant frequency measurements, one for each measure vowel.
F1	A vector of F1 formant frequency measurements, one for each measure vowel.
vowel_categories	The number of vowel categories we should consider, which translates to the number of gaussian mixture components in the GMM.
threshold	The threshold of the likelihood that the vowel formant frequency measurement must meet in order to be included in the convex hull.
center	Should the formant frequency measurements be centered using the mean frequency? This was not done in the original implementation.
scale	Should the formant frequency measurements be scaled to a 0-1 scale using the standard deviation of the formant frequencies? This was not done in the original implementation.

Details

The GMM and convex hull computations are both based on computed euclidean distances between vowels.

Value

The vowel space area (in Hz^2).

References

\insertAllCited

@examples data(pb) cVSA(pb[,"F2"],pb[,"F1"]) -> ch #Simple but informative plot plot(ch,xlab="<-Back / Front -> (F2)",ylab="<-Closed / Open -> (F1)")

FredrikKarlssonSpeech/articulated documentation built on July 20, 2023, 8:20 a.m.