SWIPE' pitch extractor software


Arturo Camacho's 2007 University of Florida dissertation [pdf] developed SWIPE', an advanced pitch extraction algorithm. Camacho's evaluation finds it superior to all other publically-available pitch trackers in a series of tasks. In a recent study, Evanini et al. (2011) report that the superiority of SWIPE' derives from its robustness to large pitch range specifications, but it at least as good as other pitch tracking algorithms even when pitch range is adapted to the speaker.

I've implemented the algorithm in C, and it runs an order of magnitude faster than the MATLAB version provided by Camacho. The current version is swipe-1.5 (click on "Downloads"), which compiles on Linux and Mac OS X. An integrated Python wrapper is also included.

This code has also been incorporated into SPTK. Camacho and colleagues have also extended this code for parallel computation to make Aud-SWIPE-P.

[ ~ ]