soundpoints: calculates acoustic landmarks for whistled sounds from an...
In landmarkacoustics/SoundPoints-R: an R version of the SoundPoints algorithm

Description Usage Arguments Value Warning Author(s) See Also Examples

each event in the sound file is represented by the same number of landmarks.

1	soundpoints(file.name, time.res, freq.res, passband, amplitude.cutoff, loudness.column, half.life, initial.amplitude, amplitude.threshold, var.columns, var.weights, n.landmarks)

`file.name`	the location of a file in WAV format
`time.res`	the desired time resolution of the spectrogram
`freq.res`	the desired frequency resolution of the spectrogram – this will be converted to the nearest power of two, based upon the sample rate of the input file
`passband`	a two-element vector that defines the frequency range within the spectrogram that will contain the landmarks
`amplitude.cutoff`	the area under the peak at each instant of the spectrogram
`loudness.column`	the variable to use when separating signal vs. noise. see `make.peaks`
`half.life`	this tunes the sensitivity of the amplitude smoother. smaller values means the moving average is more sensitive to sudden changes in amplitude. a good starting guess for this parameter is the time from one element to the next.
`initial.amplitude`	the starting amplitude of the signal vs. noise smooth. Large values exclude more noise at the beginning
`amplitude.threshold`	peaks are only counted as signal if they exceed the moving average by at least this much. If you set it at -1, for example, everything at least 10 dB quieter than the moving average will be included.
`var.columns`	the columns of the frequency contour that will be included in the landmark analysis. see `make.peaks`.
`var.weights`	when calculating the difference between successive peaks, these are the weights of each of the `var.columns`. For example, you could include time, mean frequency, and mean amplitude as your variables, and set `var.weights = 1/c(0.1,4000,Inf)`. This would cause the event detector to ignore amplitude and make a new event when successive peaks are separated by 0.1 seconds or 4 kiloHertz.
`n.landmarks`	the number of landmarks per event. Events that include fewer than `n.landmarks` peaks will be thrown out.

`file`	the name of the audio file, without path information
`spg`	the spectrogram object showing the whole file
`peaks`	the frequency contour of the whole file
`landmarks`	a data frame with one row per event, holding envelope and landmark measurements (e.g. duration, mean time, time1, time2, etc.)

This function makes a spectrogram from the WHOLE file, so don't use it on long recordings!

Benjamin N. Taft ben.taft@landmarkacoustics.com

detect.events make.peaks smooth.average plot.spg landmark.plot

##---- Should be DIRECTLY executable !! ----
##-- ==>  Define data, use random,
##--	or do  help(data=index)  for the standard data sets.

## The function is currently defined as
function (file.name, time.res, freq.res, passband, amplitude.cutoff, 
    loudness.column, half.life, initial.amplitude, amplitude.threshold, 
    var.columns, var.weights, n.landmarks) 
{
    spg <- spectrogram(file.name, time.res, freq.res)
    peaks <- make.peaks(spg, passband, amplitude.cutoff)
    peaks$smooth.amplitude <- smooth.average(peaks[, loudness.column], 
        get.time.res(spg), half.life, initial.amplitude)
    not.noise <- peaks[, loudness.column] > peaks$smooth.amplitude + 
        amplitude.threshold
    ids <- data.frame(event = rep(NA, nrow(peaks)), landmark = NA)
    ids$event[not.noise] = detect.events(peaks[not.noise, var.columns], 
        var.weights)
    ids$landmark[not.noise] <- assign.to.landmarks(ids$event[not.noise], 
        n.landmarks)
    no.smooth <- -ncol(peaks)
    landmarks <- aggregate(peaks[not.noise, no.smooth], list(event = ids$event[not.noise]), 
        mean)
    names(landmarks)[-1] <- paste("mean", names(landmarks)[-1], 
        sep = ".")
    landmarks$duration <- aggregate(peaks$time[not.noise], list(event = ids$event[not.noise]), 
        function(x) {
            diff(range(x))
        })$x
    for (j in 1:n.landmarks) {
        f <- not.noise & ids$landmark == j
        if (any(f)) {
            temp <- aggregate(peaks[f, no.smooth], list(event = ids$event[f]), 
                mean)
            names(temp)[-1] <- paste(names(temp)[-1], j, sep = ".")
            landmarks <- cbind(landmarks, temp[match(landmarks$event, 
                temp$event), -1])
        }
    }
    landmarks <- landmarks[is.finite(apply(landmarks, 1, sum)), 
        ]
    ix <- grep("time", names(landmarks))[-1]
    landmarks[, ix] <- landmarks[, ix] - landmarks$mean.time
    peaks$event <- ids$event
    peaks$landmark <- ids$landmark
    answer <- list(file = basename(file.name), spg = spg, peaks = peaks, 
        landmarks = landmarks)
    return(invisible(answer))
  }