I'm trying to script automatic noise removal using SoX. Some of the silences that are selected by the script have a brief "yeah" or "um" in them. I'm trying to figure out how the noise profile is calculated so I can see if these brief vocalizations will really effect the noise removal process or not. So, my question is: how is a "noise profile" calculated using noiseprof in SoX? Is it averaged across the silence that is selected?
According to this website, the noise is averaged, and these vocalizations shouldn't have a big effect on noise removal