#!/usr/bin/env bash
|
|
# This script operates on a data directory, such as in data/train/,
|
# and writes new segments to stdout. The file 'segments' maps from
|
# utterance to time offsets into a recording, with the format:
|
# <utterance-id> <recording-id> <segment-begin> <segment-end>
|
# This script assumes utterance and recording ids are the same (i.e., that
|
# wav.scp is indexed by utterance), and uses durations from 'utt2dur',
|
# created if necessary by get_utt2dur.sh.
|
|
. ./path.sh
|
|
if [ $# != 1 ]; then
|
echo "Usage: $0 [options] <datadir>"
|
echo "e.g.:"
|
echo " $0 data/train > data/train/segments"
|
exit 1
|
fi
|
|
data=$1
|
|
if [ ! -s $data/utt2dur ]; then
|
local/data/get_utt2dur.sh $data 1>&2 || exit 1;
|
fi
|
|
# <utt-id> <utt-id> 0 <utt-dur>
|
awk '{ print $1, $1, 0, $2 }' $data/utt2dur
|
|
exit 0
|
