nkululeko.segment

The nkululeko.segment module performs voice activity detection (VAD) and segmentation of audio into analyzable chunks.

Use Cases

  • Removing leading/trailing silence.

  • Splitting long recordings into uniform segments.

  • Improving feature consistency by focusing on voiced regions.

Invocation

python -m nkululeko.segment --config examples/exp_emodb_os_svm.ini

Typical INI Options

[DATA]
segment = vad
segment.min_duration = 0.5
segment.max_duration = 5.0
segment.pad = 0.1

Outputs

Segmented audio references or intermediate files; logs summarizing discarded silence.

Tips

  • Tune min_duration to avoid overly short, noisy segments.

  • Padding preserves phonetic boundaries when truncating.