Skip to contents

Track pitch and voiced speech using the TANDEM-STRAIGHT algorithm

trk_tandem.Rd

Estimates F0 and per-frame voicing probability using a gammatone filterbank combined with neural network-based pitch tracking (Hu & Wang 2010), which simultaneously segregates voiced speech from noise. TANDEM is robust to noise and reverberation and can track multiple simultaneous pitch sources.

Usage

trk_tandem(listOfFiles, minF = 50, maxF = 500, target_sample_rate = 20000, return_mask = FALSE, toFile = FALSE, explicitExt = "tnd", outputDirectory = NULL, verbose = TRUE, ...)

Arguments

listOfFiles

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

...

Additional arguments (currently unused).

minF

Numeric. Minimum F0 in Hz. Default 50 Hz.

maxF

Numeric. Maximum F0 in Hz. Default 500 Hz.

target_sample_rate

Numeric. Internal processing sample rate in Hz. TANDEM requires 20000 Hz. Default 20000.

return_mask

Logical. Return time-frequency voiced mask (currently unused). Default FALSE.

toFile

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

explicitExt

Character. Output file extension. Default "tnd".

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:

pitch

REAL64, fundamental frequency in Hz, n_frames × 1. Zero indicates unvoiced frames.

voicing_prob

REAL64, voicing probability, 0–1, n_frames × 1.

Frame rate: 100 Hz (fixed 10 ms hop). If toFile = TRUE: character vector of output file paths.

Note

The core processing is currently a placeholder; full TANDEM C++ integration is under development. Results reflect the algorithm framework but may not match the published TANDEM-STRAIGHT output.

References

Hu G, Wang D (2010). “A tandem algorithm for pitch estimation and voiced speech segregation.” IEEE Transactions on Audio, Speech, and Language Processing, 18(8), 2067–2079. doi:10.1109/TASL.2010.2041110 .

Hu K, Wang D (2011). “Unvoiced speech segregation from nonspeech interference via CASA and spectral subtraction.” IEEE Transactions on Audio, Speech, and Language Processing, 19(6), 1600–1609. doi:10.1109/TASL.2010.2093893 .

See also

trk_pitch_rapt, trk_pitch_swipe, trk_pitch_yin for other pitch tracking methods

Examples

if (FALSE) { # \dontrun{
# Basic pitch tracking
result <- trk_tandem("speech.wav")
plot(result$pitch, type = "l", main = "TANDEM Pitch Track")

# With noisy speech
result <- trk_tandem("noisy_speech.wav", minF = 80, maxF = 400)

# Batch processing
files <- c("speaker1.wav", "speaker2.wav", "speaker3.wav")
results <- trk_tandem(files, verbose = TRUE)

# Save to files
trk_tandem("speech.wav", toFile = TRUE, outputDirectory = "output/")
} # }