Skip to content
Latest commit 93b4a6c @MattShannon standardize tops of lots of files to clarify copyright
Ignore-this: 86e221c67ed30c9047012535e921bdee



Directory structure

ap: Aperiodicity components extracted by STRAIGHT (ascii).

cmp: Training data in HTK data format (natural endian). 
     You can see these files as HList -C hlist.conf cmp/cmu_us_arctic_slt_a0001.cmp.

f0: Fundamental frequency values extracted by STRAIGHT (ascii).

gv: GV probability density functions (4-byte float, natural endian).

labels: Context-independent and -dependent labels in HTK label format (ascii).

lf0: Log fundamental frequency values converted from F0 files in f0 directory (4-byte float, natural endian).

lists: Context-independent and -dependent model lists (ascii).

mgc: Mel-generalized cepstral (MGC) coefficients or MGC-LSPs extracted by SPTK (4-byte float, natural endian).

questions: Question files for decision tree-based context clustering (ascii).

raw: 16kHz raw audio (2-byte short, little endian).

scp: Data path files (ascii).

scripts: shell, perl, and tcl scripts (ascii).

sp: Spectra extracted by STRAIGHT. 

utts: Utterance information files in Festival utt format (ascii).

win: windows coefficients to calculate dynamic features (ascii).


This demonstration uses part of the CMU ARCTIC database, which is covered by the following copyrights:
    Copyright 2003 Carnegie Mellon University

See `License` for details of license and warranty.
Something went wrong with that request. Please try again.