soundgen (version 3.0.0)

specToMS: Spectrogram to modulation spectrum and back

Description

specToMS() takes a spectrogram (complex or magnitude) and returns its 2D Fourier transform - the modulation spectrum (MS). msToSpec() is the inverse: it takes a full, complex modulation spectrum and reconstructs the magnitude spectrogram.

Usage

specToMS(spec, center = TRUE)
msToSpec(ms)

Value

specToMS() returns a list with $ms (the uncentered MS) and $ms_centered (the centered MS, or NULL if center = FALSE). Both are complex matrices of the same dimension as spec, with spectral modulation (FM) in rows and temporal modulation (AM) in columns.

msToSpec() returns a magnitude spectrogram (real matrix, frequency in rows, time in columns).

Arguments

spec

target spectrogram (numeric matrix, frequency in rows, time in columns), with frequency labels (kHz) in rownames and time labels (ms) in colnames

center

if TRUE, the MS is centered with DC in the middle

ms

centered modulation spectrum (complex matrix, spectral modulation in rows, temporal modulation in columns), as returned by specToMS()

Details

Both functions assume that the input carries valid labels: a spectrogram must have frequency labels (kHz) in its rownames and time labels (ms) in its colnames, and a modulation spectrum must have spectral modulation labels in its rownames and temporal modulation labels in its colnames. This is the case for objects produced by soundgen functions (e.g. spectrogram, stft_simple, modulationSpectrum), and it also ensures that zero-padding is reflected correctly in the labels.

See Also

modulationSpectrum, plotMS

Examples

s = soundgen(sylLen = 200, amFreq = 25, amDep = 50, pitch = 250,
             addSilence = 0, samplingRate = 16000)
spec = spectrogram(s, samplingRate = 16000, windowLength = 25, step = 5,
  plot = FALSE)
image(t(spec))
ms = specToMS(spec)
plotMS(log(Mod(ms$ms_centered)), quantiles = NULL)
# round-trip: reconstruct the magnitude spectrogram
spec_new = msToSpec(ms$ms_centered)
image(t(spec_new))