pwelch: Estimate Power Spectral Density Using Welch's Method

In PAMmisc: Miscellaneous Functions for Passive Acoustic Analysis

Estimate Power Spectral Density Using Welch's Method

Description

Estimates the power spectral density (PSD) of an input signal using Welch's method. This should function similarly to the Matlab function pwelch, but results may not be identical. Breaks the input signal into (usually) overlapping frames and averages the resulting PSD estimates

Usage

pwelch(
  x,
  nfft,
  noverlap = 0,
  sr = NULL,
  window = NULL,
  demean = c("long", "short", "none"),
  channel = 1
)

Arguments

x	input signal, either a numeric vector, Wave, WaveMC, or audioSample object. Can also be a path to a wav file and it will be read in
nfft	length of FFT window to use for individual frames
noverlap	number of samples each frame should overlap
sr	sample rate of data, only necessary if x is a vector
window	window to apply, must be a vector of length nfft. If NULL (default), then a hamming window will be used
demean	method of demeaning the signal, one of 'long', 'short', or 'none'. Long subtracts the mean of the entire signal x, short subtracts the mean of each individual frame, none does no mean subtraction.
channel	channel number to analyse, ignored if x is a vector

Value

returns a list with items spec, the PSD estimate of the input signal, and freq, the frequency values (Hz) at each value of spec

Author(s)

Taiki Sakai [e-mail]

Examples

# wav example is synthetic echolocation clicks at 4kHz
wavFile <- system.file('extdata/testWav.wav', package='PAMmisc')
psd <- pwelch(wavFile, nfft=1e3, noverlap=500, demean='long')
plot(x=psd$freq, y=10*log10(psd$spec), type='l')

PAMmisc documentation built on April 12, 2025, 2:09 a.m.