Can you force align it and then use just the pauses? Or, I've used an R script with the tuneR package to load in wav files and detect changes in amplitude between frames. Low, unchanging energy is probably a pause and a spike in amp / high amp is probably a return to speech
-
-
-
i do have these forced aligned but the pauses from the forced aligner aren't reliable. i'm thinking a simple acoustic-based approach could help me flag files with long pauses.
Kraj razgovora
Novi razgovor -
-
-
I have a Matlab script that I can give you that allows you to mark boundaries of .wav files and saves the stat and end times in a .xlsx file. Would that be useful?
-
Nah, I have one of those. I’m trying to flag files that might have really long pauses
Kraj razgovora
Novi razgovor -
-
-
Did you try annotate to textgrid (silences)? There’s a bit to play with there.
-
That's what I've done (script to annotate > textgrid (silences)), after adjusting some of the params, and coupled with a script to manually confirm and/or adjust silence boundaries if need be. Mfa had pretty good silence boundaries for me too, but depends on your data
- Još 1 odgovor
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.
dad! data scientist studying how children with cerebral palsy talk. bayesian. cats.