I extracted eGeMAPSv02 features from 16kHz audio, and all the formant bandwidths have values of a median of >1000 (see histogram below)
I was curious if this was a configuration error, or if the units are not in Hz. It also could perhaps be an audio issue, but the audio sounds normal to me and looks fine on a spectogram. I spent some time digging through the source but I got a bit lost after a while.
Hello!
I extracted eGeMAPSv02 features from 16kHz audio, and all the formant bandwidths have values of a median of >1000 (see histogram below)
I was curious if this was a configuration error, or if the units are not in Hz. It also could perhaps be an audio issue, but the audio sounds normal to me and looks fine on a spectogram. I spent some time digging through the source but I got a bit lost after a while.
Thanks!
Rahul