getPeakFrequency: Find peak frequency

In allisonglider/seabiRds: Prepares biologging data for analysis and upload to the lab database

Find peak frequency

Description

Find peak frequency

Usage

getPeakFrequency(
  data,
  time,
  method = c("pracma", "fft"),
  window,
  frequency = NULL,
  threshold = 0.1,
  sample = 1,
  maxfreq = NULL
)

Arguments

data	Vector of numeric values for finding peaks
time	Vector of POSIXct times
method	Method for calculating peak frequency, options are 'pracma' or 'fft'
window	Size of moving window for estimating peaks, in seconds
frequency	Sampling frequency of data, in Hz, if missing the function will estimate frequency from time
threshold	Minimum amplitude to be include as a true peak, results may differ with method
sample	Only used in fft method, only calculate paeaks at sample interval, in sec, to save processing time
maxfreq	Maximum frequency of interest using the FFT method

Details

The pracma method uses pracma::findpeaks() to count the number of peaks within each window above a specified treshold. Results are returned as peaks per second. This approach is faster than the fft method and can be less impacted by noise in the data.If using this approach, I recommend exploring different threshold values to understand how this changes the result.

The fft method uses a Fast Fourier Transform to identify the peak frequency in the input data over a moving window defined the frequency * interval. The threshold paramter filters out peaks identified in samples with an inter-quartile range less than the threshold value. This helps eliminate noisy peaks with low amplitudes. This method can be slow for large data sets. Use sample to only calcualte peaks at a sampling interval (default is 1 sec), this can speed up the algorithm.

Value

A vector of frequencies (Hz).

Examples

# generate data with a noisy sine wave in the middle
ts <- 25 # sampling rate of 25 Hz
time <- 500 # duration in sec
l <- ts * time
x <- seq(0, time - 1/25, 1/25)
f <- 8 # frequency of 8Hz
f <- rep(rnorm(time, f, 0.05), each = ts) # create noise to frequency
amp <- rnorm(ts * time, 0.5, 0.1) # create noisy amplitude
z <- amp*sin(2*pi*f*x) # generate sine wave
# add data with no frequency to start and end
dat <- c(rep(0, 1500) + rnorm(1500, 0, 0.02), z, rep(0, 1500) + rnorm(1500, 0, 0.04))
tim <- seq(Sys.time(), length.out = length(dat), by = 1/ts) # generate time vector

# example using pracma
myPeaks <- getPeakFrequency(data = dat, time = tim, method = 'pracma', window = 5,
frequency = NULL, threshold = 0.1)
par(mfrow = c(2,1))
plot(dat ~ tim, type = 'l')
plot(myPeaks ~ tim, type = 'l')
par(mfrow = c(1,1))

# example using fft
myPeaks <- getPeakFrequency(data = dat, time = tim, method = 'fft', window = 5,
frequency = NULL, threshold = 0.06)
par(mfrow = c(2,1))
plot(dat ~ tim, type = 'l')
plot(myPeaks ~ tim, type = 'l')
par(mfrow = c(1,1))

allisonglider/seabiRds documentation built on Feb. 14, 2025, 7:37 a.m.