Hello! I have an HMM-DNN trained model and I have a script that creates the alignment and ctm file for a given audio file and corresponding transcription.
I am using get_train_ctm.sh to do it, but I am getting times at the utterance level. Because of this I created the segments file and a reco2file_and_channel file ("utt_id utt_id A"). The original wav file that I am working with has around 38 seconds. After creating these 2 mentioned files, I am getting the same result I was obtaining before, at the utterance level. The TRUe for segments is on. Anyone knows what can be the error?
Can I create the reco2file_and_channel like I mentioned above?
Thanks