merge_vad_map.txt
663 Bytes
# This table defines the mapping used by the binary merge-vads to
# combine the output of compute-vad and compute-vad-from-frame-likes.
# The first column corresponds to VAD decisions from compute-vad
# and the second corresponds to VAD decisions from
# compute-vad-from-frame-likes. The labels "0" and "1" in the
# first column represent (approximately) silence and nonsilence
# respectively. The labels "0," "1," and "2" in the second column
# represent noise, speech, and music, respectively. The third
# column lists the resulting output labels: "0," "1," and "2"
# corresponding to silence/noise, speech, and music.
0 0 0
1 0 0
0 1 0
1 1 1
0 2 0
1 2 2