peakTrough: Find Peaks and Troughs in a Spectrum

In TaikiSan21/PAMmisc: Miscellaneous Functions for Passive Acoustic Analysis

Find Peaks and Troughs in a Spectrum

Description

Finds up to three peaks in a spectrum, as well as the troughs between those peaks.

Usage

peakTrough(spec, freqBounds = c(10, 30), dbMin = -15, smooth = 5, plot = FALSE)

Arguments

spec	the spectrum of a signal, the first column must be frequency in kilohertz, the second column must be dB
freqBounds	a two element vector specifying the frequency range around the highest peak to search for a second/third peak. Units are in kHz, a value of c(f1, f2) requires a second peak to be at least f1 kHz away from the first peak, but no further than f2 kHz away.
dbMin	minimum dB level for second / third peaks, relative to maximum dB. Any points lower than this dB level will not be considered a candidate peak.
smooth	the amount to smooth the spectrum before attempting to find second / third peaks. Uses a simple local average, smooth is the total number of points to use. A value of 1 applies no smoothing.
plot	logical flag to plot image of peak/trough locations on spectrum. Useful for finding appropriate settings for freqBounds and dbMin

Details

The first peak is the frequency with the highest dB level (first and last frequency points are ignored). Then this uses a very simple algorithm to find second and third peaks in a spectrum. Peak candidates are identified with a few simple steps:

Step 1

Use a local average of (smooth) points to smooth the spectrum.

Step 2

Check if a point is larger than both its neighbors.

Step 3

Check if points are within the frequency range specified by freqBounds. Points must be at least f1 kHz away from the frequency , but no further than f2 kHz away.

Step 4

Check if points are above the minimum dB level specified by dbMin.

From the remaining points the point with the highest dB level is selected as the second peak, then the frequency range filter of Step 3 is applied again around this second peak before attempting to find a third peak. If no second or third peak is found (ie. no values fall within the specified frequency and dB search ranges), then it will be set to 0. The trough values are set as the frequency with the lowest dB level between any peaks that were found. The trough values will be 0 for any peaks that were not found.

If you are unsure of what levels to specify for freqBounds and dbMin, setting plot=TRUE will show a visualization of the search range and selected peaks so you can easily see if the selected parameters are capturing the behavior you want.

Value

a dataframe with the frequencies (in kHz) of up to 3 peaks and 2 troughs between those peaks. Also reports the peak-to-peak distance. Any peaks / troughs that were not able to be found (based on freqBounds and dbMin parameters) will be 0.

Author(s)

Taiki Sakai [e-mail]

Examples

clickWave <- createClickWave(signalLength = .1, clickLength = 1000, clicksPerSecond = 200,
                             frequency = 3e3, sampleRate = 10e3)
peakTrough(seewave::spec(clickWave, plot=FALSE), plot=TRUE)

TaikiSan21/PAMmisc documentation built on April 6, 2024, 5:33 p.m.