Tool(s) to automate finding sleep talking in 8h MP3 files?

[Corvinian]

Hi,

I am recording my sleep and have ca. 8 hours of recording per day. Recording only upon voice activation has a too high thresdhold I cannot change in the digital recorder, so I cannot use this. Now I cannot listen to 8 hours of sleep multiplied by xx days.

Audacity came to my mind as first stop.
Is there some Plugin or tool I can use:
- to find sound, label it (the Silence/Sound Finder)?
- to read timestamp and calculate the current time right before next dream, log to a file
(later if possible: record the timestamp as sound and insert in audio, too)
- automate this for hundreds of files.

[billw58]

Sound Finder might be able to do the job, but it may have the same problem as the voice activation on the recorder - false positives from non-speech noises. You'll have to experiment to find the right settings.

Audacity does not record time stamps, so you'll have to use the time line and know when the recording was started.

Once you have all the talking segments labelled you can do File > Export Labels, then bring the resulting text file into a spreadsheet.

There's no way I know of to automate this process in Audacity.

-- Bill

[Trebor]

it's possible to display the whole 8 hours on one screen, then zoom in on peaks in the waveform which may be the somniloquy [ooh!] you are looking for, (or a cat's chorus).

click on the magnifying glass icon on the far right to display the whole [X hour] recording on the screen ...


http://forum.audacityteam.org/viewtopic ... 25#p152225

depending on you computer you may have to wait a minute for Audacity to redraw the waveform when you zoom in on part of a recording which is several hours long, for zooming in see ... http://forum.audacityteam.org/viewtopic ... 25#p152225

An automatic solution would be "truncate silence" which would remove all the recording which was below a threshold : cutting out the boring "silence" ("silence" = sound below user set threshold). However I've never used "truncate silence" on a recording which was more than a few minutes long. The time line would be lost after "truncate silence" : you would not know when the somniloquy occurred.

[Corvinian]

Just learned the real term for sleep talking: somniloquy
Thank you for the good ideas, ... and a term to "dig" the web.

Has anybody tried automating some Audacity-tasks, with AutoHotKey, for example ?
Can AHK be used to do this kind of work, with the 'truncate silence' feature on each file?

[Trebor]

I've not tried batch processing files in Audacity, apparently it's not designed for that ...

Re: (How) Can i batch convert files with audacity ?

Postby steve » Thu Mar 18, 2010 3:22 am
I'd not be using Audacity for batch converting files.
Yes it can be done, but it's like banging in nails with a shoe.

http://forum.audacityteam.org/viewtopic ... 071#p78071


BTW "truncate silence", if it can cope with a file 8 hours long, is going to take 10 or 15 minutes [guesstimate] to process a file of that length, depending on the speed of your computer.

[waxcylinder]

Has anybody tried automating some Audacity-tasks, with AutoHotKey, for example ?

With any luck you might get hot-keys for Effects (and the Generate and remaining Analyze menu items) in the next release of Audacity 2.0.1.

One of the developers has been working on this and Gale & I have been testing it for him - it's in the current alpha test code and seems to work well. It's still windows only at the moment.

The ones I find really useful are the Fade-In and Fade-Out effects with hot-keys.

WC

[Trebor]

Just tried Audacity “truncate silence” on a 5hour 15min mp3 (sample rate 22050Hz)
It managed it in eight minutes, and reduced its length to under an hour ,
so my guesstimate of 10 -15 minutes for 8 hours seems about right.