Skip to contents

Track formants using Time-Varying Weighted Linear Prediction (TVWLP)

trk_formant_tvwlp.Rd

Estimates formant frequencies and bandwidths using GCI-anchored quasi-closed phase (QCP) weighted, time-varying linear prediction at 8 kHz. TVWLP reduces glottal source contamination compared to standard LPC, yielding more stable F1 estimates on voiced speech. Prefer this over Burg-method trackers when F1 tracking in the closed phase is the priority.

Usage

trk_formant_tvwlp(listOfFiles, beginTime = 0, endTime = 0, windowShift = 10, npeaks = 3L, p = 8L, q = 3L, preemp = 0.97, lptype = "tvwlp_l2", toFile = TRUE, explicitExt = "tvf", outputDirectory = NULL, verbose = TRUE)

Arguments

listOfFiles

Character vector of audio file paths. Any format supported by av is accepted; audio is resampled to 8 kHz internally.

beginTime

Numeric. Start of analysis window in seconds. Default 0 (file start).

endTime

Numeric. End of analysis window in seconds. Default 0 (file end).

windowShift

Numeric. Frame shift in milliseconds; sets output frame rate (1000 / windowShift Hz). Default 10.0 ms.

npeaks

Integer. Number of formants to extract from LP roots. Default 3.

p

Integer. LPC order; must exceed 2 × npeaks for reliable root extraction. Default 8.

q

Integer. Polynomial degree for the time-varying LP coefficients (controls within-segment variation). Default 3.

preemp

Numeric. Pre-emphasis coefficient (0–1). Default 0.97.

lptype

Character. LP solver method: "tvwlp_l2" (GCI-weighted L2, default), "tvlp_l2" (unweighted L2), "tvwlp_l1" (weighted L1, requires quantreg), "tvlp_l1" (unweighted L1).

toFile

Logical. If TRUE, write SSFF output files and return the count written invisibly. If FALSE, return an AsspDataObj. Default TRUE.

explicitExt

Character. Output file extension. Default "tvf".

outputDirectory

Character. Directory for output files. NULL (default) writes alongside the input file.

verbose

Logical. Print per-file progress. Default TRUE.

Value

If toFile = FALSE: an AsspDataObj with tracks:

fm

REAL32, formant frequencies in Hz, n_frames × npeaks. Columns correspond to F1, F2, …, F{npeaks}.

bw

REAL32, formant bandwidths in Hz, n_frames × npeaks.

Frame rate: 1000 / windowShift Hz (default 100 Hz). If toFile = TRUE: integer count of files written, returned invisibly.

Details

The weighted methods (tvwlp_l2, tvwlp_l1) internally estimate pitch via SRH and GCIs via SEDREAMS at 8 kHz. The unweighted variants skip pitch estimation and are faster but less accurate on voiced speech. A 5-frame median filter is applied to each formant track before output.

References

(El-Jaroudi and Makhoul 1991)

Examples

if (FALSE) { # \dontrun{
# Default 3-formant TVWLP tracking
res <- trk_formant_tvwlp("recording.wav", toFile = FALSE)
dim(res$fm)  # n_frames x 3

# Unweighted LP (no pitch/GCI estimation)
trk_formant_tvwlp("speech.mp3", lptype = "tvlp_l2")
} # }