Skip to contents

Measure acoustic parameters in batches of sound files

Source: R/spectro_analysis.R
spectro_analysis.Rd

spectro_analysis measures acoustic parameters on acoustic signals for which the start and end times are provided.

Usage

spectro_analysis(
  X,
  bp = "frange",
  wl = 512,
  wl.freq = NULL,
  threshold = 15,
  parallel = 1,
  fast = TRUE,
  path = NULL,
  pb = TRUE,
  ovlp = 50,
  wn = "hanning",
  fsmooth = 0.1,
  harmonicity = FALSE,
  nharmonics = 3,
  ...
)

Arguments

X

'selection_table', 'extended_selection_table' or data frame with the following columns: 1) "sound.files": name of the sound files, 2) "sel": number of the selections, 3) "start": start time of selections, 4) "end": end time of selections.

bp

A numeric vector of length 2 for the lower and upper limits of a frequency bandpass filter (in kHz) or "frange" (default) to indicate that values in bottom.freq and top.freq columns will be used as bandpass limits. Lower limit of bandpass filter is not applied to fundamental frequencies.

wl

A numeric vector of length 1 specifying the spectrogram window length. Default is 512. See 'wl.freq' for setting windows length independently in the frequency domain.

wl.freq

A numeric vector of length 1 specifying the window length of the spectrogram for measurements on the frequency spectrum. Default is 512. Higher values would provide more accurate measurements. Note that this allows to increase measurement precision independently in the time and frequency domain. If NULL (default) then the 'wl' value is used.

threshold

amplitude threshold (%) for fundamental frequency and dominant frequency detection. Default is 15.

parallel

Numeric. Controls whether parallel computing is applied. It specifies the number of cores to be used. Default is 1 (i.e. no parallel computing).

fast

Logical. If TRUE (default) then the peakf acoustic parameter (see below) is not computed, which substantially increases performance (~9 times faster). This argument will be removed in future version.

path

Character string containing the directory path where the sound files are located. If NULL (default) then the current working directory is used.

pb

Logical argument to control progress bar and messages. Default is TRUE.

ovlp

Numeric vector of length 1 specifying % of overlap between two consecutive windows, used for fundamental frequency (using fund or FF) and dominant frequency (using dfreq). Default is 50.

wn

Character vector of length 1 specifying window name. Default is hanning'. See function ftwindow for more options.

fsmooth

A numeric vector of length 1 to smooth the frequency spectrum with a mean sliding window (in kHz) used for mean peak frequency detection. This help to average amplitude "hills" to minimize the effect of amplitude modulation. Default is 0.1.

harmonicity

Logical. If TRUE harmonicity related parameters (fundamental frequency parameters [meanfun, minfun, maxfun], hn_freq, hn_width, harmonics and HNR) are measured. Note that measuring these parameters considerably increases computing time.

nharmonics

Numeric vector of length 1 setting the number of harmonics to analyze.

...

Additional parameters to be passed to analyze, which measures parameters related to harmonicity.

Value

Data frame with 'sound.files' and 'selec' as in the input data frame, plus the following acoustic parameters:

  • duration: length of signal (in s)

  • meanfreq: mean frequency (in kHz). Calculated as the weighted average of the frequency spectrum (i.e. weighted by the amplitude within the supplied band pass).

  • sd: standard deviation of frequency (in kHz). Calculated as the weighted standard deviation of the frequency spectrum (i.e. weighted by the amplitude within the supplied band pass).

  • freq.median: median frequency. The frequency at which the frequency spectrum is divided in two frequency intervals of equal energy (in kHz)

  • freq.Q25: first quartile frequency. The frequency at which the frequency spectrum is divided in two frequency intervals of 25% and 75% energy respectively (in kHz)

  • freq.Q75: third quartile frequency. The frequency at which the frequency spectrum is divided in two frequency intervals of 75% and 25% energy respectively (in kHz)

  • freq.IQR: interquartile frequency range. Frequency range between 'freq.Q25' and 'freq.Q75' (in kHz)

  • time.median: median time. The time at which the time envelope is divided in two time intervals of equal energy (in s)

  • time.Q25: first quartile time. The time at which the time envelope is divided in two time intervals of 25% and 75% energy respectively (in s). See acoustat

  • time.Q75: third quartile time. The time at which the time envelope is divided in two time intervals of 75% and 25% energy respectively (in s). See acoustat

  • time.IQR: interquartile time range. Time range between 'time.Q25' and 'time.Q75' (in s). See acoustat

  • skew: skewness. Asymmetry of the frequency spectrum (see note in specprop description)

  • kurt: kurtosis. Peakedness of the frequency spectrum (see note in specprop description)

  • sp.ent: spectral entropy. Energy distribution of the frequency spectrum. Pure tone ~ 0; noisy ~ 1. See sh

  • time.ent: time entropy. Energy distribution on the time envelope. ~0 means amplitude concentrated in a specific time point, 1 means amplitude equally distributed across time. See th

  • entropy: spectrographic entropy. Product of time and spectral entropy sp.ent * time.ent. See H

  • sfm: spectral flatness. Similar to sp.ent (Pure tone ~ 0; noisy ~ 1). See sfm

  • meandom: average of dominant frequency measured across the spectrogram

  • mindom: minimum of dominant frequency measured across the spectrogram

  • maxdom: maximum of dominant frequency measured across the spectrogram

  • dfrange: range of dominant frequency measured across the spectrogram

  • modindx: modulation index. Calculated as the cumulative absolute difference between adjacent measurements of dominant frequencies divided by the dominant frequency range (measured on the spectrogram). 1 means the signal is not modulated.

  • startdom: dominant frequency measurement at the start of the signal (measured on the spectrogram).

  • enddom: dominant frequency measurement at the end of the signal(measured on the spectrogram).

  • dfslope: slope of the change in dominant frequency (measured on the spectrogram) through time ((enddom-startdom)/duration). Units are kHz/s.

  • peakf: peak frequency. Frequency with the highest energy. This parameter can take a considerable amount of time to measure. It's only generated if fast = FALSE. It provides a more accurate measure of peak frequency than 'meanpeakf' but can be more easily affected by background noise. Measured on the frequency spectrum.

  • meanpeakf: mean peak frequency. Frequency with highest energy from the mean frequency spectrum (see meanspec). Typically more consistent than peakf in the presence of noise.

  • meanfun: average of fundamental frequency measured across the acoustic signal. Only measured if harmonicity = TRUE.

  • minfun: minimum fundamental frequency measured across the acoustic signal. Only measured if harmonicity = TRUE.

  • maxfun: maximum fundamental frequency measured across the acoustic signal. Only measured if harmonicity = TRUE.

  • hn_freq: mean frequency of the 'n' upper harmonics (kHz) (see analyze). Number of harmonics is defined with the argument 'nharmonics'. Only measured if harmonicity = TRUE.

  • hn_width: mean bandwidth of the 'n' upper harmonics (kHz) (see analyze). Number of harmonics is defined with the argument 'nharmonics'. Only measured if harmonicity = TRUE.

  • harmonics: the amount of energy in upper harmonics, namely the ratio of total spectral power above 1.25 x F0 to the total spectral power below 1.25 x F0 (dB) (see analyze). Number of harmonics is defined with the argument 'nharmonics'. Only measured if harmonicity = TRUE.

  • HNR: harmonics-to-noise ratio (dB). A measure of the harmonic content generated by getPitchAutocor. Only measured if harmonicity = TRUE.

Details

The function measures 29 acoustic parameters (if fast = TRUE) on each selection in the data frame. Most parameters are produced internally by specprop, fpeaks, fund, and dfreq from the package seewave and analyze from the package soundgen. NAs are produced for fundamental and dominant frequency measures when there are no amplitude values above the threshold. Additional parameters can be provided to the internal function analyze, which measures parameters related to harmonicity.

References

Araya-Salas, M., & Smith-Vidaurre, G. (2017). warbleR: An R package to streamline analysis of animal acoustic signals. Methods in Ecology and Evolution, 8(2), 184-191.

Author

Marcelo Araya-Salas (marcelo.araya@ucr.ac.cr) and Grace Smith Vidaurre

Examples

{
data(list = c("Phae.long1", "Phae.long2", "Phae.long3", "lbh_selec_table"))
writeWave(Phae.long1, file.path(tempdir(), "Phae.long1.wav"))
writeWave(Phae.long2, file.path(tempdir(), "Phae.long2.wav"))
writeWave(Phae.long3, file.path(tempdir(), "Phae.long3.wav"))

# measure acoustic parameters
sp_param <- spectro_analysis(X = lbh_selec_table[1:8,], pb = FALSE, path = tempdir())

# measuring peakf
sp_param <- spectro_analysis(X = lbh_selec_table[1:8,], pb = FALSE, fast = FALSE, path = tempdir())

# \donttest{
# measuring harmonic-related parameters using progress bar
sp_param <- spectro_analysis(X = lbh_selec_table[1:8,], harmonicity = TRUE, 
path = tempdir(), ovlp = 70)
# }
}