This package contains codes for sonority-based syllabification of speech signals.
Version 0.21 (5.10.2017)
(please see https://github.com/speech-utcluj/thetaOscillator-syllable-segmentation for Adriana Stan's Python implementation of the syllabifier)
Basic pipeline: 1) compute gammatone-envelopes for speech signals with 1000 Hz sampling rate 2) call thetaOscillator.m with the Gammatone envelopes as input 3) Syllable boundaries are in the "bounds" and "bounds_t" output variables (in 1000 Hz frames and seconds, respectively).
Adjust threshold parameter of thetaOscillator for sensitivity adjustments.
See SylSegDemo.m for an example.
(c) Okko Rasanen, [email protected] , 2016.
If you use this algorithm in publications, please cite:
Rasanen, O., Doyle, G. & Frank, M. C. (2018). Pre-linguistic segmentation of speech into syllable-like units. Cognition, 171, 130–150.
Note that the package uses Gammatone-filterbank front-end by Ning Ma (http://staffwww.dcs.shef.ac.uk/people/N.Ma/resources/gammatone/). If you are not using 64-bit OS X environment, compile the function first with "mex gammatone_c.c". Also, re-compiling might be required for newer versions of 64-bit OS X is the computation freezes without an apparent reason.
Also uses peakdet.m from Eli Billauer (see the file for license).