compare.methods: Assessing the performance of acoustic distance measurements

Description

compare.methods creates graphs to visually assess performance of acoustic distance measurements

Usage

compare.methods(X = NULL, flim = c(0, 22), bp = c(0, 22), mar = 0.1, wl = 512, ovlp = 90, 
res = 150, n = 10, length.out = 30, methods = c("XCORR", "dfDTW", "ffDTW", "SP"), 
it = "jpeg", parallel = 1, path = NULL, sp = NULL, pb = TRUE, grid = TRUE, 
clip.edges = TRUE, threshold = 15, na.rm = FALSE, scale = FALSE,
 pal = reverse.gray.colors.2, img = TRUE, ...)

Arguments

'selection.table' object or data frame with results from manualoc function, autodetec function, or any 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 c(0, 22).

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 c(0, 22). Note that for XCORR this argument sets the frange argument from the xcorr function.

mar

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

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.

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: cross-correlation (XCORR, from xcorr), dynamic time warping on dominant frequency time series (dfDTW, from dtw applied on dfts output), dynamic time warping on dominant frequency time series (ffDTW, from dtw applied on ffts output), spectral parameters (SP, from specan).

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.

Data frame with acoustic parameters as the one generated by specan. Must contain 'sound.files' and "selec' columns and the same selections as in 'X'.

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 remainging 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 specreator).

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 xcorr), dynamic time warping on dominant frequency time series (dfDTW, from dtw applied on dfts output), dynamic time warping on dominant frequency time series (ffDTW, from dtw applied on ffts output), spectral parameters (SP, from specan). The graph also contains 2 scatterplots (1 for each method) of the acoustic space of all signals in the input data frame 'X'. 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 (princomp) is applied to calculate distances when using spectral parameters (SP). In that case the first 2 PC's are used. Classical Multidimensional Scalling (also known as Principal Coordinates Analysis, (cmdscale)) is used for all other methods. Note that SP can only be used with at least 22 selections (number of rows in input data frame) as PCA only works with more units than variables. The graphs are return as image files in the working directory. The file name contains the methods being compared and the rownumber of the selections. This function uses internally a modified version of the spectro function from seewave package to create spectrograms.

Examples

Run this code

# NOT RUN {
# Set temporary working directory
# setwd(tempdir())

data(list = c("Phae.long1", "Phae.long2", "Phae.long3", "Phae.long4", "selec.table"))
writeWave(Phae.long1,"Phae.long1.wav")
writeWave(Phae.long2,"Phae.long2.wav")
writeWave(Phae.long3,"Phae.long3.wav")
writeWave(Phae.long4,"Phae.long4.wav")

compare.methods(X = 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")

#remove progress bar
compare.methods(X = 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)

#check this folder!
getwd()


#compare SP and XCORR
#first we need to create a larger data set as the PCA that summarizes the spectral parameters
#needs more units (rows) that variables (columns)
#so I just create a new selection table repeating 3 times selec.table
st2 <- rbind(selec.table, selec.table, selec.table)

#note that the selection labels should be also changed
st2$selec <- 1:nrow(st2)
#now we can compare SP method against XCORR
compare.methods(X = st2, 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")

#compare SP method against dfDTW
compare.methods(X = st2, 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")

#alternatively we can provide our own SP matrix
sp <- specan(selec.table, bp = c(0, 10))

#and selec just a few variables to avoid the problem of # observations vs # parameters in PCA
sp <- sp[, 1:7]

compare.methods(X = selec.table, flim = c(0, 10), sp = sp, 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")

#note that "SP" should also be included as a method in 'methods'
#again, all images are saved in the working directory
# }
# NOT RUN {
# }