Skip to contents

Extract Multiple Parts from a Sound

Source: R/vad.R
sound_extract_parts.Rd

Extracts multiple time intervals from a Sound object and returns them as a list of Sound objects. Useful for extracting voiced segments identified by voice activity detection.

Usage

sound_extract_parts(
  sound,
  start_times,
  end_times,
  window_shape = "rectangular",
  relative_width = 1,
  preserve_times = FALSE,
  return_r6 = TRUE
)

Arguments

sound

Sound object

start_times

Numeric vector of interval start times (seconds)

end_times

Numeric vector of interval end times (seconds)

window_shape

Character. Window shape for extraction (default: "rectangular"). See details for all options.

relative_width

Numeric. Relative width of window (default: 1.0). For gaussian2/kaiser2, use 2.0. For gaussian3-5, use 3.0-5.0 respectively.

preserve_times

Logical. Preserve original time stamps (default: FALSE)

return_r6

Logical. Return R6 Sound objects (TRUE) or raw xptrs (FALSE). Using FALSE skips R6 wrapper construction.

Value

List of Sound objects, one for each extracted interval

Details

This function is vectorized to extract multiple intervals efficiently. Each extracted sound can then be concatenated or analyzed separately.

Available window shapes (see Praat manual for details): - "rectangular" (default) - No tapering - "triangular" - Triangular (Bartlett) taper - "parabolic" - Parabolic (Welch) taper - "hanning" - Hanning window - "hamming" - Hamming window - "gaussian1" - Gaussian window (sd=0.42466) - "gaussian2" - Narrower Gaussian (sd=0.21233), use relative_width=2.0 - "gaussian3" - Even narrower (sd=0.14155), use relative_width=3.0 - "gaussian4" - Very narrow (sd=0.10616), use relative_width=4.0 - "gaussian5" - Extremely narrow (sd=0.08493), use relative_width=5.0 - "kaiser1" - Kaiser-Bessel window (alpha=20.7) - "kaiser2" - Narrower Kaiser-Bessel (alpha=40.5), use relative_width=2.0

References

Praat documentation: https://www.fon.hum.uva.nl/praat/manual/Sound__Extract_part___.html

Examples

sound <- sounds_append(
  Sound$create_tone(frequency = 200, duration = 0.5, amplitude = 0.8),
  Sound$create_tone(frequency = 200, duration = 0.3, amplitude = 0.001)
)
vad_grid <- sound_to_textgrid_silences(sound)
voiced_intervals <- textgrid_get_intervals_where(vad_grid, 1, "equals",
 "sounding")

voiced_sounds <- sound_extract_parts(
  sound,
  voiced_intervals$xmin,
  voiced_intervals$xmax
)

# Analyze each segment separately
for (i in seq_along(voiced_sounds)) {
cat("Segment", i, "duration:", voiced_sounds[[i]]$get_total_duration(),
 "s\n")
}
#> Segment 1 duration: 0.52 s