seewave (version 2.2.3)

spectro: 2D-spectrogram of a time wave

Description

This function returns a two-dimension spectrographic representation of a time wave. The function corresponds to short-term Fourier transform. An amplitude contour plot can be overlaid.

Usage

spectro(wave, f, channel = 1, wl = 512, wn = "hanning", zp = 0,
ovlp = 0, noisereduction = NULL, fastdisp = FALSE,
complex = FALSE, norm = TRUE, correction="none",
fftw = FALSE, dB = "max0", dBref = NULL, plot = TRUE,
flog = FALSE, grid = TRUE, osc = FALSE, scale = TRUE, cont = FALSE,
collevels = NULL, palette = spectro.colors,
contlevels = NULL, colcont = "black",
colbg = "white", colgrid = "black",
colaxis = "black", collab="black",
cexlab = 1, cexaxis = 1, 
tlab = "Time (s)",
flab = "Frequency (kHz)",
alab = "Amplitude",
scalelab = "Amplitude\n(dB)",
main = NULL,
scalefontlab = 1, scalecexlab =0.75,
axisX = TRUE, axisY = TRUE, tlim = NULL, trel = TRUE,
flim = NULL, flimd = NULL,
widths = c(6,1), heights = c(3,1),
oma = rep(0,4),
listen=FALSE,
...)

Value

This function returns a list of three items:

time

a numeric vector corresponding to the time axis.

freq

a numeric vector corresponding to the frequency axis.

amp

a numeric or a complex matrix corresponding to the amplitude values. Each column is a Fourier transform of length wl/2.

Arguments

wave

an R object.

f

sampling frequency of wave (in Hz). Does not need to be specified if embedded in wave.

channel

channel of the R object, by default left channel (1).

wl

window length for the analysis (even number of points) (by default = 512).

wn

window name, see ftwindow (by default "hanning").

zp

zero-padding (even number of points), see Details.

ovlp

overlap between two successive windows (in %).

noisereduction

a numeric vector of length 1, if 1 a noise reduction is applied along the rows of the spectrogram, if 2 a noise reduction applied along the columns. See Details.

fastdisp

faster graphic display for long wave. The spectrogram/oscillogram is displayed/saved faster in the graphic device/ graphic file when set to TRUE, with a cost on graphical resolution.

complex

if TRUE the STFT will be returned as complex numbers.

norm

if TRUE the STFT is normalised (i. e. scaled) by its maximum.

correction

a character vector of length 1 to apply an amplitude ("amplitude") or an energy ("energy") correction to each FT window. This argument is useful only when one wish to obtain absolute values that is when norm=FALSE. By default no correction is applied ("none").

fftw

if TRUE calls the function FFT of the library fftw. See Notes.

dB

a character string specifying the type dB to return: "max0" (default) for a maximum dB value at 0, "A", "B", "C", "D", and "ITU" for common dB weights. If set to NULL, then a linear scale is used.

dBref

a dB reference value. NULL by default but should be set to 2*10e-5 for a 20 microPa reference.

plot

logical, if TRUE plots the spectrogram (by default TRUE).

flog

a logical to plot the frequency on a logarithmic scale.

grid

logical, if TRUE plots a y-axis grid (by default TRUE).

osc

logical, if TRUE plots an oscillogram beneath the spectrogram (by default FALSE).

scale

logical, if TRUE plots a dB colour scale on the right side of the spectrogram (by default TRUE).

cont

logical, if TRUE overplots contour lines on the spectrogram (by default FALSE).

collevels

a set of levels which are used to partition the amplitude range of the spectrogram (in dB).

palette

a color palette function to be used to assign colors in the plot, see Details.

contlevels

a set of levels which are used to partition the amplitude range for contour overplot (in dB).

colcont

colour for cont plotting.

colbg

background colour.

colgrid

colour for grid plotting.

colaxis

color of the axes.

collab

color of the labels.

cexlab

size of the labels.

cexaxis

size of the axes.

tlab

label of the time axis.

flab

label of the frequency axis.

alab

label of the amplitude axis.

scalelab

amplitude scale label.

main

label of the main title.

scalefontlab

font of the amplitude scale label.

scalecexlab

cex of the amplitude scale label.

axisX

logical, if TRUE plots time X-axis (by default TRUE).

axisY

logical, if TRUE plots frequency Y-axis (by default TRUE).

tlim

modifications of the time X-axis limits.

trel

time X-axis with a relative scale when tlim is not null, i.e. relative to wave.

flim

modifications of the frequency Y-axis limits (in kHz).

flimd

dynamic modifications of the frequency Y-axis limits. New wl and ovlp arguments are applied to increase time/frequency resolution.

widths

a vector of length 2 to control the relative widths of columns on the device when scale is TRUE.

heights

a vector of length 2 to control the relative heights of rows on the device when osc is TRUE.

oma

a vector of length 4 to control the size of outer margins when either scale or osc is TRUE.

listen

if TRUE the sound is played back (by default FALSE).

...

other contour and oscillo graphical parameters.

Author

Jerome Sueur and Caroline Simonis.

Details

Following Heisenberg uncertainty principle, the short-term Fourier transform cannot be precised in both time and frequency. The temporal and frequency precisions of the function are actually dependent of the wl value. Choosing a high wl value will increase the frequency resolution but reduce the temporal one, and vice versa. The frequency precision is obtained by calculating the ratio f/wl, and the temporal precision is obtained by calculating the reverse ratio wl/f. This problem can be reduced in some way with zp that adds 0 values on both sides of the analysis window. This increases frequency resolution without altering time resolution.
Any colour palette can be used. In particular, it is possible to use other palettes coming with seewave: temp.colors, reverse.gray.colors.1, reverse.gray.colors.2, reverse.heat.colors, reverse.terrain.colors, reverse.topo.colors, reverse.cm.colors corresponding to the reverse of heat.colors, terrain.colors, topo.colors, cm.colors.
Use locator to identify points. The noise reduction using the argument noisereduction is an image filter, not a signal filter. The principle consists in subtracting each spectrogram row or column by its median. Noise reduction alters energy conservation, it should then be used for visual display only.

References

Hopp, S. L., Owren, M. J. and Evans, C. S. (Eds) 1998. Animal acoustic communication. Springer, Berlin, Heidelberg.

See Also

ggspectro, spectro3D, lts, dynspec, wf, oscillo, dBscale, fft.

Examples

if (FALSE) {
data(tico)
data(pellucens)
# simple plots
spectro(tico,f=22050)
spectro(tico,f=22050,osc=TRUE)
spectro(tico,f=22050,scale=FALSE)
spectro(tico,f=22050,osc=TRUE,scale=FALSE)
# change the dB scale by setting a different dB reference value (20microPa)
spectro(tico,f=22050, dBref=2*10e-5)
# unnormalised spectrogram with a linear amplitude scale
spectro(tico, dB=NULL, norm=FALSE, scale=FALSE)
# manipulating wl
op<-par(mfrow=c(2,2))
spectro(tico,f=22050,wl=256,scale=FALSE)
title("wl = 256")
spectro(tico,f=22050,wl=512,scale=FALSE)
title("wl = 512")
spectro(tico,f=22050,wl=1024,scale=FALSE)
title("wl = 1024")
spectro(tico,f=22050,wl=4096,scale=FALSE)
title("wl = 4096")
par(op)
# vertical zoom using flim
spectro(tico,f=22050, flim=c(2,6))
spectro(tico,f=22050, flimd=c(2,6))
# a full plot
pellu2<-cutw(pellucens,f=22050,from=1,plot=FALSE)
spectro(pellu2,f=22050,ovlp=85,zp=16,osc=TRUE,
    cont=TRUE,contlevels=seq(-30,0,20),colcont="red",
    lwd=1.5,lty=2,palette=reverse.terrain.colors)
# black and white spectrogram 
spectro(pellu2,f=22050,ovlp=85,zp=16,
    palette=reverse.gray.colors.1)
# colour modifications
data(sheep)
spectro(sheep,f=8000,palette=temp.colors,collevels=seq(-115,0,1))
spectro(pellu2,f=22050,ovlp=85,zp=16,
palette=reverse.cm.colors,osc=TRUE,colwave="orchid1") 
spectro(pellu2,f=22050,ovlp=85,zp=16,osc=TRUE,palette=reverse.heat.colors,
colbg="black",colgrid="white", colwave="white",colaxis="white",collab="white")
}