norm {emuR}

Normalise speech data

Description

Normalises speech data

Usage

norm(data, speakerlabs, type = "gerst", rescale = FALSE)

Arguments

data

A matrix of data. Can be either an n-columned matrix or a trackdata object as returned by track.

speakerlabs

A parallel vector of speaker labels.

type

The type of extrinsic normalisation to be performed on data. type can be "nearey", "cen", "lob", "gerst" (default), for normalisation according to Nearey, centroid method, Lobanov, or Gerstman.

rescale

Currently only works for Lobanov normalisation. The normalised values are multiplied by the standard deviation and then the mean is added, where the standard deviation and mean are across all original speakers' unnormalised data.

Details

Types of normalisation: "nearey", Nearey : Find the log of each data element and subtract from each the mean of the logarithmic data. "cen", centroid: Find the mean of the data column and subtract it from each data element in that column. "lob", Lobanov: Find the mean and standard deviation of the data. Subtract the mean from each data element and divide each result by the standard deviation. "gerst", Gerstman: Subtract from the data the minimum formant value then divide by the formant range.

Value

Normalised values of data are returned, having the same structure as data.

See Also

track