Skip to contents

Assessing the performance of acoustic distance measurements

Source: R/compare_methods.R
compare_methods.Rd

compare_methods creates graphs to visually assess performance of acoustic distance measurements

Usage

compare_methods(
  X = NULL,
  flim = NULL,
  bp = NULL,
  mar = 0.1,
  wl = 512,
  ovlp = 90,
  res = 150,
  n = 10,
  length.out = 30,
  methods = NULL,
  it = "jpeg",
  parallel = 1,
  path = NULL,
  custom1 = NULL,
  custom2 = NULL,
  pb = TRUE,
  grid = TRUE,
  clip.edges = TRUE,
  threshold = 15,
  na.rm = FALSE,
  scale = FALSE,
  pal = reverse.gray.colors.2,
  img = TRUE,
  ...
)

Arguments

X

'selection_table' object or data frame with columns for sound file name (sound.files), selection number (selec), and start and end time of signal (start and end). Default NULL.

flim

A numeric vector of length 2 for the frequency limit in kHz of the spectrogram, as in spectro. Default is NULL.

bp

numeric vector of length 2 giving the lower and upper limits of the frequency bandpass filter (in kHz) used in the acoustic distance methods. Default is NULL.

mar

Numeric vector of length 1. Specifies plot margins around selection in seconds. Default is 0.1.

wl

A numeric vector of length 1 specifying the window length of the spectrogram and cross-correlation, default is 512.

ovlp

Numeric vector of length 1 specifying the percent overlap between two consecutive windows, as in spectro. Default is 90.

res

Numeric argument of length 1. Controls image resolution. Default is 150.

n

Numeric argument of length 1. Defines the number of plots to be produce. Default is 10.

length.out

A character vector of length 1 giving the number of measurements of fundamental or dominant frequency desired (the length of the time series). Default is 30.

methods

A character vector of length 2 giving the names of the acoustic distance methods that would be compared. The methods available are:

  • XCORR: cross-correlation (cross_correlation function)

  • dfDTW: dynamic time warping on dominant frequency contours (freq_DTW function)

  • ffDTW: dynamic time warping on fundamental frequency contours (freq_DTW function)

  • SP: spectral parameters (spectro_analysis function)

  • SPharm: spectral parameters (spectro_analysis function with argument harmonicity = TRUE)

  • MFCC: statistical descriptors of Mel frequency cepstral coefficients (mfcc_stats function)

Default NULL.

it

A character vector of length 1 giving the image type to be used. Currently only "tiff" and "jpeg" are admitted. Default is "jpeg".

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).

path

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

custom1

Data frame containing user parameters. The data frame must have 4 columns: the first 2 columns are 'sound.files' and "selec' columns as in 'X', the other 2 (columns 3 and 4) are 2 numeric columns to be used as the 2 parameters representing custom measurements. If the data has more than 2 parameters try using PCA (i.e. prcomp function)to summarize it in 2 dimensions before using it as an input. Default is NULL.

custom2

Data frame containing user parameters with the same format as 'custom1'. 'custom1' must be provided first. Default is NULL.

pb

Logical argument to control progress bar. Default is TRUE.

grid

Logical argument to control the presence of a grid on the spectrograms (default is TRUE).

clip.edges

Logical argument to control whether edges (start or end of signal) in which amplitude values above the threshold were not detected will be removed when using dfDTW and ffDTW methods. If TRUE this edges will be excluded and signal contour will be calculated on the remaining values. Default is TRUE.

threshold

amplitude threshold (%) for dominant and/or fundamental frequency detection when using dfDTW, ffDTW and SP methods. Default is 15.

na.rm

Logical. If TRUE all NAs produced when pairwise cross-correlations failed are removed from the results. This means that all selections with at least 1 cross-correlation that failed are excluded in both methods under comparison. Only apply if XCORR is one of the methods being compared.

scale

Logical. If TRUE dominant and/or fundamental frequency values are z-transformed using the scale function, which "ignores" differences in absolute frequencies between the signals in order to focus the comparison in the frequency contour, regardless of the pitch of signals. Default is TRUE.

pal

A color palette function to be used to assign colors in the spectrograms, as in spectro. Default is reverse.gray.colors.2.

img

A logical argument specifying whether an image files would be produced. Default is TRUE.

...

Additional arguments to be passed to a modified version of spectro for customizing graphical output. This includes fast.spec, an argument that speeds up the plotting of spectrograms (see description in spectrograms).

Value

Image files with 4 spectrograms of the selection being compared and scatterplots of the acoustic space of all signals in the input data frame 'X'.

Details

This function produces graphs with spectrograms from 4 signals in the provided data frame that allow visual inspection of the performance of acoustic distance methods at comparing those signals. The signals are randomly picked up from the provided data frame (X argument).The spectrograms are all plotted with the same frequency and time scales. The function compares 2 methods at a time. The methods available are: cross-correlation (XCORR, from cross_correlation), dynamic time warping on dominant frequency time series (dfDTW, from dtw applied on freq_ts output), dynamic time warping on dominant frequency time series (ffDTW, from dtw applied on freq_ts output), spectral parameters (SP, from spectro_analysis). The graph also contains 2 scatterplots (1 for each method) of the acoustic space of all signals in the input data frame 'X', including the centroid as black dot. The compared selections are randomly picked up from the pool of selections in the input data frame. The argument 'n' defines the number of comparisons (i.e. graphs) to be produced. The acoustic pairwise distance between signals is shown next to the arrows linking them. The font color of a distance value correspond to the font color of the method that generated it, as shown in the scatterplots. Distances are standardized, being 0 the distance of a signal to itself and 1 the farthest pairwise distance in the pool of signals. Principal Component Analysis (prcomp) is applied to calculate distances when using spectral parameters (SP) and descriptors of cepstral coefficients (MFCC). In those cases the first 2 PC's are used. Classical Multidimensional Scalling (also known as Principal Coordinates Analysis, (cmdscale)) is used for cross-correlation (XCORR) and any dynamic time warping method. The graphs are return as image files in the working directory. The file name contains the methods being compared and the row number of the selections. This function uses internally a modified version of the spectro function from seewave package to create spectrograms. Custom data can also be compared against the available methods (or against each other) using the arguments 'custom1' and 'custom2'.

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.

See also

catalog

Author

Marcelo Araya-Salas (marcelo.araya@ucr.ac.cr). It uses internally a modified version of the spectro function from seewave package to create spectrograms.

Examples

if (FALSE) { # \dontrun{
# Save to temporary working directory
data(list = c("Phae.long1", "Phae.long2", "Phae.long3", "Phae.long4", "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"))
writeWave(Phae.long4, file.path(tempdir(), "Phae.long4.wav"))

compare_methods(
  X = lbh_selec_table, flim = c(0, 10), bp = c(0, 10), mar = 0.1, wl = 300,
  ovlp = 90, res = 200, n = 10, length.out = 30,
  methods = c("XCORR", "dfDTW"), parallel = 1, it = "jpeg", path = tempdir()
)

# remove progress bar
compare_methods(
  X = lbh_selec_table, flim = c(0, 10), bp = c(0, 10), mar = 0.1, wl = 300,
  ovlp = 90, res = 200, n = 10, length.out = 30,
  methods = c("XCORR", "dfDTW"), parallel = 1, it = "jpeg", pb = FALSE, path = tempdir()
)

# check this folder!
getwd()


# compare SP and XCORR
compare_methods(
  X = lbh_selec_table, flim = c(0, 10), bp = c(0, 10), mar = 0.1, wl = 300,
  ovlp = 90, res = 200, n = 10, length.out = 30,
  methods = c("XCORR", "SP"), parallel = 1, it = "jpeg", path = tempdir()
)

# compare SP method against dfDTW
compare_methods(
  X = lbh_selec_table, flim = c(0, 10), bp = c(0, 10), mar = 0.1, wl = 300,
  ovlp = 90, res = 200, n = 10, length.out = 30,
  methods = c("dfDTW", "SP"), parallel = 1, it = "jpeg",
  path = tempdir()
)

# alternatively we can provide our own SP matrix
Y <- spectro_analysis(lbh_selec_table, path = tempdir())

# selec a subset of variables
Y <- Y[, 1:7]

# PCA
Y <- prcomp(Y[, 3:ncol(Y)])$x

# add sound files and selec columns
Y <- data.frame(lbh_selec_table[, c(1, 3)], Y[, 1:2])

compare_methods(
  X = lbh_selec_table, methods = c("dfDTW"), custom1 = Y,
  path = tempdir()
)
} # }