speech-segmentation-transcription

Detect the child speech segments based on the manually annotated mom speech segments, modify mom speech segment intervals, and output both to textgrids. Export both detected mom and child wav segments, and transcribe them using kaldi ASPIRE model. Write the transcription results to json file.

Usage:

$ python run.py --child_puzzle_wav=
--mom_puzzle_wav=
--mom_puzzle_textgrid=
--child_outfile_textgrid=
--child_segment_wav_outdir=
--mom_segment_wav_outdir=
--add_seconds_at_boundary=\

(detailed info in docstring)

Example result:

(example input: MCRP_ID#3001_G1.wav and MCRP_ID#3001_G1.TextGrid, apply extract_session.py first to get the puzzle session audios)

egs.TextGrid: output textgrid of child and mom speech segments detedcted by the program
JSONData.json: json file of transcriptions by the program
text_whole: easy visualization of dialog between child and mom, transcribed by the program
child_audio_trans_by_eesen: transcription on audio recorded by mic on child using speech kitchen EESEN
mom_audio_trans_by_eesen: transcription on audio recorded by mic on mom using speech kitchen EESEN