Post A5oEDpyDpeGNI1k4eW by cadadr@mastodon.sdf.org
 (DIR) More posts by cadadr@mastodon.sdf.org
 (DIR) Post #A5oEDpyDpeGNI1k4eW by cadadr@mastodon.sdf.org
       2021-04-01T17:11:19Z
       
       0 likes, 0 repeats
       
       Reading https://ling.auf.net/lingbuzz/005790 was somewhat disappointing.Apparently when recording with phone or Zoom (the meeting app) F0 is fairly intact, F1-2 less so, and F3 is unreliable. Also, Zoom spoils intensity data.These results are from participants that were linguists, some of which were phoneticians, and were recorded in "quiet home conditions".This is worrisome... I need the intensity data. And while F0 variation is not significant the outliers seem to be consequential..
       
 (DIR) Post #A5oEDqSi0IeKoa0QAi by publius@mastodon.sdf.org
       2021-04-01T17:19:41Z
       
       0 likes, 0 repeats
       
       @cadadr I'm pretty sure the data compression model Zoom uses for audio incorporates certain assumptions about speech, so that what you receive could justifiably be described as "partly synthesized". Hence for your purposes I would expect such audio to be worse than useless.
       
 (DIR) Post #A5oEgYHEh77KoSKOPo by cadadr@mastodon.sdf.org
       2021-04-01T17:24:54Z
       
       0 likes, 0 repeats
       
       @publius That might be why the weird random intensity drops are happening in their data. Apparently during random intervals the intensity measurements drop extremely, very visibly in the spectra. Data from women is also qualitatively but obviously worse.Even if that wasn't the case I think there's an ethical case against relying on Zoom for research...