This reminds me of the handheld gizmo that paranormal investigators on the show Ghost Adventures have that records inaudible frequencies and converts them into words in an electronic voice!
On a visit to the castle of Vlad the Impaler (Dracula):
Investigator Aaron: "Who's there?"
Elephants also speak in complex low vibration rumbles that travel for miles but can't be heard by the naked human ear.
In regard to non-verbal cues: though RealSense does not have a facial emotion recognition component any more, there are ways to read a face. For example, I analyze the angle of a facial landmark such as the mouth or eyebrow to determine whether it is angled up (happy) or angled down (angry).