Skip to contents

Measure frequency spectrum correlation

Source: R/spectrum_correlation.R
spectrum_correlation.Rd

spectrum_correlation() measures frequency spectrum correlation of sounds referenced in an extended selection table. Spectral correlation measures the similarity of two sounds in the frequency domain.

Usage

spectrum_correlation(
  X,
  cores = getOption("mc.cores", 1),
  pb = getOption("pb", TRUE),
  cor.method = c("pearson", "spearman", "kendall"),
  spec.smooth = getOption("spec.smooth", 5),
  hop.size = getOption("hop.size", 11.6),
  wl = getOption("wl", NULL),
  ovlp = getOption("ovlp", 70),
  path = getOption("sound.files.path", "."),
  n.bins = 100
)

Arguments

X

The output of set_reference_sounds(), an object of class data.frame, selection_table, or extended_selection_table (the last 2 classes are created by warbleR::selection_table() from the warbleR package) with the test sound files' annotations. Must contain the following columns: 1) sound.files: name of the .wav files, 2) selec: unique selection identifier (within a sound file), 3) start: start time and 4) end: end time of selections, 5) bottom.freq: low frequency for bandpass, 6) top.freq: high frequency for bandpass, 7) sound.id: ID of sounds used to identify counterparts across distances, and 8) reference: identity of sounds to be used as reference for each test sound (row). See set_reference_sounds() for more details on the structure of X.

cores

Numeric vector of length 1. Controls whether parallel computing is applied by specifying the number of cores to be used. Default 1 (i.e. no parallel computing). Can be set globally for the current R session via the "mc.cores" option (see options()).

pb

Logical argument to control if progress bar is shown. Default TRUE. Can be set globally for the current R session via the "pb" option (see options()).

cor.method

Character string indicating the correlation coefficient to be applied ("pearson", "spearman", or "kendall", see stats::cor()).

spec.smooth

Numeric vector of length 1 determining the length of the sliding window used for a sum smooth for power spectrum calculation (in kHz). Default 5.

hop.size

Numeric vector of length 1 specifying the time window duration (in ms). Default 11.6 ms, which is equivalent to 512 wl for a 44.1 kHz sampling rate. Ignored if wl is supplied. Can be set globally for the current R session via the "hop.size" option (see options()).

wl

A vector with a single even integer number specifying the window length of the spectrogram. Default NULL. If supplied, hop.size is ignored. Odd integers will be rounded up to the nearest even number. Can be set globally for the current R session via the "wl" option (see options()).

ovlp

Numeric vector of length 1 specifying the percentage of overlap between two consecutive windows, as in seewave::spectro(). Default 70. Can be set globally for the current R session via the "ovlp" option (see options()).

path

Character string containing the directory path where the sound files are found. Only needed when X is not an extended selection table. If not supplied the current working directory is used. Can be set globally for the current R session via the "sound.files.path" option (see options()).

n.bins

Numeric vector of length 1 specifying the number of frequency bins to use for representing power spectra. Default 100. If NULL, the raw power spectrum is used (note that this can result in high RAM memory usage for large data sets). Power spectrum values are interpolated using stats::approx().

Value

Object X with an additional column, spectrum.correlation, containing the computed frequency spectrum correlation coefficients.

Details

The function measures the spectral correlation coefficients of sounds in which a reference playback has been re-recorded at increasing distances. Values range from 1 (identical frequency spectrum, i.e. no degradation) to 0. The sound.id column must be used to tell the function to only compare sounds belonging to the same category (e.g. song-types). The function will then compare each sound to the corresponding reference sound. Two methods for computing spectral correlation are provided (see the method argument). The function uses seewave::meanspec() internally to compute power spectra. Use spectrum_blur_ratio() to extract raw spectra values. NA is returned if at least one of the power spectra cannot be computed.

References

Araya-Salas, M., Grabarczyk, E. E., Quiroz-Oliva, M., Garcia-Rodriguez, A., & Rico-Guevara, A. (2025). Quantifying degradation in animal acoustic signals with the R package baRulho. Methods in Ecology and Evolution, 00, 1-12. https://doi.org/10.1111/2041-210X.14481 Apol, C.A., Sturdy, C.B. & Proppe, D.S. (2017). Seasonal variability in habitat structure may have shaped acoustic signals and repertoires in the black-capped and boreal chickadees. Evol Ecol. 32:57-74.

See also

envelope_correlation() and spectrum_blur_ratio().

Other quantify degradation: blur_ratio(), detection_distance(), envelope_correlation(), plot_blur_ratio(), plot_degradation(), set_reference_sounds(), signal_to_noise_ratio(), spcc(), spectrum_blur_ratio(), tail_to_signal_ratio()

Author

Marcelo Araya-Salas (marcelo.araya@ucr.ac.cr)

Examples

{
  # load example data
  data("test_sounds_est")

  # method 1
  # add reference column
  Y <- set_reference_sounds(X = test_sounds_est)

  # run spectrum correlation
  spectrum_correlation(X = Y)

  # method 2
  Y <- set_reference_sounds(X = test_sounds_est, method = 2)
  # spectrum_correlation(X = Y)
}