[Audacity-devel] EffectTruncSilence almost working ... newbie requesting help to finish

[Lynn A. <l_d...@ad...>]

<alert comment="newbie">

I've uploaded TruncSilenceEffect.zip to:
http://osdn.dl.sourceforge.net/sourceforge/inverse/TruncSilenceEffect.zip

I "claim" it is about 90+% working as far as truncating long silent
gaps in mono vocal tracks (but that may be naive on my part :-)

It is "cloned" from Dominic's EffectNormalize code.

The .zip includes the following files:
* TruncSilence.cpp
* TruncSilence.h
* LoadEffects.cpp
* audacity.dsw ... Visual Studio project file to reference above files

These files are in a directory structure that should unzip into the
standard Audacity 1.2.3 directories:
Audacity
... src
... src\effects\TruncSilence.cpp
... src\effects\TruncSilence.h
... src\effects\LoadEffects.cpp
... win\audacity.dsp

There is a .wav file input.wav in
http://osdn.dl.sourceforge.net/sourceforge/inverse/TruncSilence.zip
that can be used to test the effect. It is about 30 seconds long, with
quite a few silent gaps.

Caveats:
**************
* Developed using Visual Studio 98 (Visual C++ 6.0 sp6). The
equivalent of makefiles for other platforms would need to be adjusted
to incorporate TruncSilence.cpp and TruncSilence.h
**************
* Assumes mono track that has been normalized and had noise removed.
Otherwise, there may be residual noise that will appear to be a signal
that should be removed retain. (Click/pop removal would be very
helpful when that is eventually implemented.)
**************
* Dialog to acquire parameters is "stubbed". It should use wxWidget
text boxes to acquire parameters for:
- what length of time is considered silence (current code is
"hard-wired" to 0.2 sec of continuous signal strength beneath a 0.01
threshold),

- a parameter as to what db is considered silence (current code is
"hard-wired to 0.01, which is about -40db),

- a parameter as to what length of time is considered speech
(something under 0.01 might be considered a click/pop so don't reset
the counter for a long silence ... not implemented at all in the code
... mea culpa).

- I have effectively zilch wxWidget experience, so was hoping to get
help from someone with LOTS more wxWidget experience than myself. TIA!
**************
* I'm baffled by how tracks are made up of sequences and/or blocks.
The current code detects silent gaps over 0.2 seconds, and correctly
replaces them with a 0.2 sec of silence. But my attempts to use:
WaveTrack::cut
or
track->GetEnvelope()->CollapseRegion
were less than successful. The mono track retains some of the "right
hand side" that should be removed.

(e.g.  a track has 5 seconds of silence, 3 sec of continuous speech, 5
sec of silence, 8 sec of speech, 2 sec of silence = 23 seconds
overall. The three long silent gaps are each replaced by 0.2 sec of
silence. The speech is retained and "pulled to the left" correctly.
But instead of the overall track being 0.2 + 3.0 + 0.2 + 8.0 + 0.2 =
11.6 seconds, it is about 15 seconds long. It is shortened in overall
duration, but not sufficiently. There is still unaltered speech on the
"right hand side" of the remaining track.)
**************
At this point, I would appreciate assistance from someone with LOTS
more audacity experience than myself regarding sequences and blocks.
TIA!

Let me know if you have questions, and my apologies for residual
deficiencies in the code. I probably should have stared at it longer,
and scratched my head until I got it working, but the code seems to be
in a state where someone who knows what he/she is doing can get it
working a LOT faster than I can, and in such a way to be more
consistent with the rest of Audacity.

</alert>