How to use the Silence Recognizer MPI-PL in ELAN
The Silence Recognizer:
- segments a wav file into silence and non-silence chunks
- works locally, there's no need to be online
The Silence Recognizer segments a wav file into silence and non-silence chunks on the basis of a user specified threshold.
Non-silence does not necessary refer to speech, it can be any kind of sound above the background noise level.
The user interface contains the following elements:
- Files list: here the audio file can be selected that will be processed. The list contains all wav
files that are linked to the transcription.
- Silence level: provides two ways of specifying the silence/non-silence threshold:
- Select manually: provides a slider to specify the silence level in dB relative to the peak RMS (Root Mean Square). The level value is also shown as a percentage.
- Select by example: allows to add selections (either created by hand or taken from a tier) that represent silence (the background noise). The threshold is calculated based on the samples in the selections.
In case of stereo wav files it is possible to add different selections per channel, the resulting segmentation will differ per channel then.
- Settings panel
- Minimal silence duration: allows to specify a minimal duration for silence segments. Shorter segments will be filtered out.
- Minimal non-silence duration: allows to specify a minimal duration for non-silence segments. Shorter segments will be filtered out.