Skip to contents

LPC

Source: R/lpc-wrapper.R
LPC.Rd

Praat LPC object for linear predictive coding analysis, created via direct C++ module binding.

Arguments

.xptr

Not for direct use. External pointer to the underlying C++ LPC object; set internally when a method returns a new LPC.

Value

An LPC object with methods for linear predictive coding analysis and inverse filtering.

Details

LPC (Linear Predictive Coding) estimates the spectral envelope of a sound by modeling it as an autoregressive process. The LPC coefficients describe the vocal tract filter and can be converted to formants, spectra, or other representations.

Creating LPC objects

LPC objects are created from Sound objects using one of several methods:

  • sound$to_lpc_burg() - Burg method (fastest, most robust)

  • sound$to_lpc_auto() - autocorrelation method

  • sound$to_lpc_covariance() - covariance method

  • sound$to_lpc_marple() - Marple method (slowest, most accurate)

Query methods

  • get_number_of_frames() - number of analysis frames

  • get_time_step() - time step between frames

  • get_sampling_period() - sampling period of the original sound

  • get_max_num_coefficients() - maximum number of LPC coefficients

  • get_gain_at_frame(frame) - gain value for a specific frame

  • get_coefficients_at_frame(frame) - LPC coefficients for a specific frame

  • get_all_gains() - vector of all gain values

  • get_all_coefficients() - matrix of all LPC coefficients

Conversion methods

  • to_formant(margin) - not available in this build (requires CLAPACK); use Sound$to_formant_burg() for formant extraction instead

  • to_spectrum(time, ...) - convert to a Spectrum at a specific time

  • to_matrix() - convert to a Matrix object

Voice source extraction (inverse filtering)

  • filter_inverse(sound) - extract glottal flow by inverse filtering

  • filter_inverse_at_time(sound, time, channel) - use the filter from a specific time

These methods remove vocal tract resonances to reveal the voice source (glottal flow waveform), useful for voice quality research and vocal fold dynamics.

See also

Sound, Formant, Spectrum, LFCC

Examples

# Load sound
sound <- Sound$create_tone(frequency = 150, duration = 0.3)

# Compute LPC (Burg method is recommended)
lpc <- sound$to_lpc_burg(
  prediction_order = 16,
  analysis_width = 0.025,
  time_step = 0.005,
  pre_emphasis_frequency = 50.0
)

# Query properties
n_frames <- lpc$get_number_of_frames()
gains <- lpc$get_all_gains()
coeffs <- lpc$get_all_coefficients()

# Get coefficients for a specific frame
coef_frame1 <- lpc$get_coefficients_at_frame(1)

# Convert to other representations
spectrum <- lpc$to_spectrum(time = 0.15, df_min = 20)

# Extract voice source (glottal flow) via inverse filtering at a given time
midpoint <- sound$get_duration() / 2
glottal_flow <- lpc$filter_inverse_at_time(sound, time = midpoint)