Integrate a data checker/gatekeeper

[felixbur]

I keep having problems with datasets that contain wav files that are

• not 16khz mono wav
• zero length
• no speech contained
• too short

would be nice to have a flag that checks the data before processing (i.e. train and devel) and removes faulty ones

[felixbur]

probably easiest to use a VAD for this, e.g. Inaspeechsegmenter

[felixbur]

or a very simple approach of VAD like here https://github.com/marsbroshok/VAD-python

[felixbur]

or a very simple approach of VAD like here https://github.com/marsbroshok/VAD-python

i tried that and it didn't really work: using some test samples with street noise and such it always declared a lot of speech

[felixbur]

started with version 0.53.0
added
check_size
check_vad (with silero)

not tested though....

[felixbur]

should add check_samplerate