— Audio annotation with Praat
Our main efforts so far on audio analysis has been focused on our automated audio processing pipeline. However, there are also interesting possibilities for annotating audio manually using Praat, as well as using Praat for feature extraction and search. We invite users of Praat to add examples and use cases.
Speech rate detection
To explore audio features and to annotate the corpus with audio information, Praat has a wealth of capabilities. Here is a working example from speechrate, run against a wav file created using clip on a video file. The arguments are Show text grid no, Silence threshold -25, Minimum dip between peaks 3 dB, Minimum pause duration 0.3 seconds:
csa@cartago:~/praat$ praat syllableCount.praat no -25 3 0.3 wav
soundname, nsyll, npause, dur (s), phonationtime (s), speechrate (nsyll/dur), articulation rate (nsyll / phonationtime), ASD (speakingtime/nsyll)
2014-12-17_2200_US_AlJazeera_Inside_Story, 6883, 266, 1792.51, 1653.18, 3.84, 4.16, 0.240
So that works -- 6883 syllables, 266 pauses, 1792.51 seconds, 1653.18 phonation time, 3.84 syllables/second speech rate, 4.16 syllables/second articulation rate (?), 0.240 speaking time per syllable. Not what we want of course, but it demonstrates we're not that far away.