Skip to contents

Pharyngeal Voice Quality Analysis

lst_pharyngeal.Rd

Extract pharyngealization/voice quality measures from labeled vowel intervals. Analyzes spectral measures (H1-H2, H1-A1, H1-A2, H1-A3, etc.) at vowel onset and midpoint with Iseli & Alwan (2004) normalization.

Usage

lst_pharyngeal(listOfFiles, textgridPath = NULL, intervalTier = 3, intervalNumber = 1, beginTime = NULL, endTime = NULL, minPitchInitial = 50, maxPitchInitial = 800, toFile = FALSE, return_jstf = FALSE, explicitExt = "pha", outputDirectory = NULL, verbose = TRUE)

Arguments

listOfFiles

Character vector of file paths to audio files (WAV, MP3, etc.)

textgridPath

Character vector of TextGrid file paths (same length as listOfFiles)

intervalTier

Integer specifying which TextGrid tier contains the vowel intervals (default: 3)

intervalNumber

Integer specifying which interval to analyze per file (default: 1)

beginTime

Numeric vector of start times in seconds (optional, overrides TextGrid if provided)

endTime

Numeric vector of end times in seconds (optional, overrides TextGrid if provided)

minPitchInitial

Minimum pitch for initial pitch detection in Hz (default: 50)

maxPitchInitial

Maximum pitch for initial pitch detection in Hz (default: 800)

toFile

Logical - write results to JSTF file (default: FALSE)

explicitExt

File extension for output files (default: "pha")

outputDirectory

Directory for output files (default: NULL = same as input)

verbose

Logical - show progress messages (default: TRUE)

return_jstf

Logical. Return JsonTrackObj instead of data.frame? Default FALSE. When both toFile and return_jstf are TRUE, the file is written AND the object returned.

Value

Data frame with 68 columns per file:

file

File path

start_time, mid_time, end_time, duration_ms

Timing information

f0_start, f0_mid

F0 at onset and midpoint (Hz)

f1/f2/f3_start, f1/f2/f3_mid

Formant frequencies at onset/mid (Hz)

bw1/bw2/bw3_start, bw1/bw2/bw3_mid

Formant bandwidths (Hz)

bw1/bw2/bw3_start_norm, bw1/bw2/bw3_mid_norm

Normalized bandwidths (Hz)

intensity_start, intensity_mid

Intensity at onset/mid (dB)

h1/h2_onset, h1/h2_mid

Harmonic amplitudes (dB)

h1_hz/h2_hz_onset, h1_hz/h2_hz_mid

Harmonic frequencies (Hz)

h1/h2_onset_norm, h1/h2_mid_norm

Normalized harmonic amplitudes (dB)

a1/a2/a3_onset, a1/a2/a3_mid

Formant peak amplitudes (dB)

a1_hz/a2_hz/a3_hz_onset, a1_hz/a2_hz/a3_hz_mid

Formant peak frequencies (Hz)

a3_onset_norm, a3_mid_norm

Normalized A3 amplitudes (dB)

h1_minus_h2/h1_minus_a1/h1_minus_a2/h1_minus_a3_onset

Raw differences at onset (dB)

h1_minus_h2/h1_minus_a1/h1_minus_a2/h1_minus_a3_onset_norm

Normalized differences at onset (dB)

a1_minus_a2/a1_minus_a3/a2_minus_a3_onset

Additional raw differences at onset (dB)

a1_minus_a3_onset_norm, a2_minus_a3_onset_norm

Additional normalized differences at onset (dB)

...

Same structure repeated for midpoint (*_mid)

Details

Algorithm (based on scriptPharyFullV4.praat):

  1. Two-pass adaptive pitch detection (speaker-specific F0 range)

  2. Find intensity maxima at onset/midpoint

  3. Extract formants F1/F2/F3 with bandwidths

  4. Extract 40ms Kaiser2 window at onset

  5. Create spectrum with pre-emphasis

  6. Find H1, H2 harmonics near F0, 2×F0

  7. Find A1, A2, A3 formant peaks near F1/F2/F3

  8. Apply Iseli & Alwan (2004) normalization

  9. Calculate differences (raw and normalized)

  10. If duration > 120ms, repeat for midpoint

Key Measures:

  • H1-H2: Open quotient indicator (higher = breathier)

  • H1-A1: Voice quality indicator (higher = breathier)

  • H1-A2: Spectral tilt measure

  • H1-A3: High-frequency energy indicator

  • Normalized (*_norm): Corrected for formant influence (Iseli & Alwan 2004)

Typical Values:

  • H1-H2: -10 to +10 dB

  • H1-A1: -5 to +5 dB

  • H1-A3: Often negative (A3 > H1)

Performance: ~24ms per vowel (15.7x faster than v4.8.14)

Dependencies

Requires pladdrr package (>= 4.8.16)

References

(Iseli and Alwan 2004)

See also

lst_vq, trk_praatsauce, lst_voice_report

Examples

if (FALSE) { # \dontrun{
# Analyze labeled vowels from TextGrid
results <- lst_pharyngeal(
  listOfFiles = "speech.wav",
  textgridPath = "speech.TextGrid",
  intervalTier = 3,
  intervalNumber = 1
)

# Analyze specific time ranges (no TextGrid)
results <- lst_pharyngeal(
  listOfFiles = c("vowel1.wav", "vowel2.wav"),
  beginTime = c(0.5, 1.0),
  endTime = c(0.7, 1.3)
)

# Write to JSTF files
lst_pharyngeal(
  listOfFiles = "speech.wav",
  textgridPath = "speech.TextGrid",
  toFile = TRUE
)  # Creates speech.pha

# Access key measures
cat(sprintf("H1-H2* (onset): %.2f dB\\n", results$h1_minus_h2_onset_norm[1]))
cat(sprintf("H1-A1* (onset): %.2f dB\\n", results$h1_minus_a1_onset_norm[1]))
} # }