Extract part of Sound with optional windowing — sound_extract

Extracts time range from Sound, applying optional window function. Implements Praat's "Sound: Extract part..." with full window shape support.

Usage

sound_extract_part(
  sound,
  t1,
  t2,
  window_shape = 1L,
  relative_width = 1,
  preserve_times = FALSE
)

Arguments

sound: Sound object
t1: Start time (seconds)
t2: End time (seconds)
window_shape: Window shape code or name: * 0 or "rectangular" - Rectangular (no tapering) * 1 or "triangular" - Triangular (Bartlett) * 2 or "parabolic" - Parabolic (Welch) * 3 or "hanning" - Hanning * 4 or "hamming" - Hamming * 5 or "gaussian1" - Gaussian with sd=0.42466 (relative to duration) * 6 or "gaussian2" - Gaussian with sd=0.21233 (narrower, use relative_width=2.0) * 7 or "gaussian3" - Gaussian with sd=0.14155 (use relative_width=3.0) * 8 or "gaussian4" - Gaussian with sd=0.10616 (use relative_width=4.0) * 9 or "gaussian5" - Gaussian with sd=0.08493 (use relative_width=5.0) * 10 or "kaiser1" - Kaiser-Bessel with alpha=20.7 * 11 or "kaiser2" - Kaiser-Bessel with alpha=40.5 (use relative_width=2.0)
relative_width: Relative width for windowing (default: 1.0). For gaussian2/kaiser2, use 2.0 to maintain effective window duration. For gaussian3, use 3.0. For gaussian4, use 4.0. For gaussian5, use 5.0. This extends physical extraction beyond [t1,t2] while keeping effective duration.
preserve_times: If TRUE, preserve original time domain (result spans t1 to t2). If FALSE, time-shift result to start at 0.

Value

New Sound object with extracted and windowed portion

Details

Window shapes with higher numbers (gaussian2-5, kaiser2) have narrower effective windows. To maintain comparable effective duration to gaussian1/kaiser1, use relative_width > 1.0, which extracts a longer physical segment while applying a more aggressive taper.

For spectral analysis, Kaiser2 and Gaussian2 with relative_width=2.0 are commonly used (e.g., in Praat's "To Spectrogram..." and "To Pitch (ac)... Very accurate").

References

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

Examples

sound <- Sound$create_tone(frequency = 220, duration = 3.0)

# Rectangular window (no tapering)
rect <- sound_extract_part(sound, 1.0, 2.0, window_shape = 0L)

# Gaussian1 window (standard)
gauss1 <- sound_extract_part(sound, 1.0, 2.0, window_shape = 5L,
 relative_width = 1.0)

# Gaussian2 with wider physical extraction
gauss2 <- sound_extract_part(sound, 1.0, 2.0, window_shape = 6L,
 relative_width = 2.0)

# Kaiser2 for spectral analysis
kaiser <- sound_extract_part(sound, 1.0, 2.0, window_shape = 11L,
 relative_width = 2.0)