Maybe some combination of: comparing average recent audio levels and detecting a sudden spike in the average human speech frequency range, using far field mic

Maybe some combination of: comparing average recent audio levels and detecting a sudden spike in the average human speech frequency range, using far field mic tech these companies have been focusing on to check against direction the audio source is coming from (ie, only focus on someone right in front of you), maybe auto adding the first name of the phone owner to a watch word ala “hey Siri”, etc I’m not an expert at all on the “how”, just thought could be a useful “what” that solves some friction with current tech (similar to how these automatically adjust noise cancellation settings vs needing to set manually as with competitors)