Function create_segment

scripts/eval/get_eval_set.py:512–558

Create audio segment from source audio based on timing information. Extracts a time-based segment from a source audio file and saves it as a separate WAV file. This is used to create individual utterance clips from longer CHiME-6 conversation recordings.

(src_dir, dst_dir, seg_dict)

            return int((h * 3600 + m * 60 + s) * 1000)

        def create_segment(src_dir, dst_dir, seg_dict):
            """Create audio segment from source audio based on timing information.

            Extracts a time-based segment from a source audio file and saves it
            as a separate WAV file. This is used to create individual utterance
            clips from longer CHiME-6 conversation recordings.

            Args:
                src_dir (str): Directory containing source audio files
                dst_dir (str): Directory where segmented clips will be saved
                seg_dict (dict): Dictionary containing segmentation metadata with keys:
                    - "audio_file": Source audio filename
                    - "audio_seg_file": Output segment filename
                    - "start_time": Segment start time (HH:MM:SS format)
                    - "end_time": Segment end time (HH:MM:SS format)

            Returns:
                str: Path to the created audio segment file

            Raises:
                FileNotFoundError: If source audio file doesn't exist
                OSError: If unable to create destination directory or write file
                ValueError: If timestamp format is invalid

            Example:
                >>> seg_dict = {
                ...     "audio_file": "S02_U06.wav",
                ...     "audio_seg_file": "S02_U06_0012500_0025000.wav",
                ...     "start_time": "00:00:12.5",
                ...     "end_time": "00:00:25.0"
                ... }
                >>> segment_path = create_segment("audio/", "segments/", seg_dict)

            Note:
                Creates destination directory if it doesn't exist. Uses pydub
                for audio processing, which supports various audio formats.
            """
            audio_file = os.path.join(src_dir, seg_dict["audio_file"])
            segment_file = os.path.join(dst_dir, seg_dict["audio_seg_file"])

            os.makedirs(dst_dir, exist_ok=True)
            audio = AudioSegment.from_wav(audio_file)
            start_time = timestamp_to_ms(seg_dict["start_time"])
            end_time = timestamp_to_ms(seg_dict["end_time"])
            clip = audio[start_time:end_time]
            clip.export(segment_file, format="wav")
            return segment_file

        def parallel_create_segment(args):
            """Wrapper function for parallel processing of audio segmentation.