soundgen (version 2.7.0)

getRough: Calculate roughness from modulation spectrum

Description

Internal soundgen function

Usage

getRough(m, roughRange = c(30, 150), roughMean = NULL, roughSD = NULL)

Value

Returns roughness in percent.

Arguments

m

numeric matrix of non-negative values with colnames giving temporal modulation frequency

roughRange

the range of temporal modulation frequencies that constitute the "roughness" zone, Hz

roughMean, roughSD

the mean (Hz) and standard deviation (semitones) of a lognormal distribution used to weight roughness estimates. If either is null, roughness is calculated simply as the proportion of spectrum within roughRange. If both roughMean and roughRange are defined, weights outside roughRange are set to 0; a very large SD (a flat weighting function) gives the same result as just roughRange without any weighting (see examples)

Details

Helper function for calculating roughness - the proportion of energy / amplitude in the roughness range

Examples

m = modulationSpectrum(soundgen(jitterDep = 2, addSilence = 0),
  samplingRate = 16000)$original
# proportion within roughRange
plot(soundgen:::getRough(m, roughRange = c(30, Inf))[, 1:2])
plot(soundgen:::getRough(m, roughRange = c(30, 150))[, 1:2])

# lognormal weighting function instead of roughRange
plot(soundgen:::getRough(m, roughRange = NULL,
  roughMean = 75, roughSD = 1)[, 1:2])  # narrow
plot(soundgen:::getRough(m, roughRange = NULL,
  roughMean = 75, roughSD = 5000)[, 1:2])  # very broad

# lognormal weighting function truncated at roughRange
plot(soundgen:::getRough(m, roughRange = c(30, 150),
  roughMean = 75, roughSD = 3)[, 1:2])
plot(soundgen:::getRough(m, roughRange = c(30, 150),
  roughMean = 75, roughSD = 5000)[, 1:2])
# approaches proportion in roughRange as SD --> Inf

# a nice plot weighting by amplitude
r2 = soundgen:::getRough(m)
plot(r2$freq, r2$rough, cex = r2$amp ^ 2 + .25)
sum(r2$rough)  # simple sum across all bands
sum(r2$rough * r2$amp / sum(r2$amp))  # amplitude-weighted mean
# log-amplitude-weighted mean
sum(r2$rough * log(r2$amp+1e-6) / sum(log(r2$amp+1e-6)))