Databases

Laryngograph-derived F0 values for four databases

DB1 was produced by Y. Atake. Our lx-derived reference F0 estimates are here
DB2 was produced by P. Bagshaw. It is available here. Our lx-derived reference F0 estimates are here
DB3 was produced by Nathalie Henrich, Christophe D'Alessandro, Michèle Castellango and Tuan Vu-Ngoc. Our lx-derived reference F0 estimates are here
DB4 was produced by Nick Campbell. Our lx-derived reference F0 estimates are here

For each database is provided the reference F0 estimate derived from the laryngograph signal. For each there are two files: DBx_lxf0, which is the F0 estimate, and DBx_mask, which is a list of pairs of numbers (ascii) indicating the start and stop of portions for which the estimates were judged reliable. The masks were produced by visual inspection of the laryngograph signal and F0 estimate. Estimates outside the intervals specified by the masks are not expected to be reliable. The estimate files represent F0 in Hz in 32-bit 'float' format (produced on a linux PC) at the same sampling resolution as the laryngograph waveform (16 kHz for DB1, DB3, DB4, 20 kHz for DB2). The beginning and end of the estimate files are padded with NaNs to ensure that they have the same size as the waveform files they are derived from.