So we can have a different frequency response just above/below the
audible range with the two cables? The loudspeaker/amplifier response
can be non-linear in this range and so transfer energy into the audible
range. The source can be chosen to have plenty of energy just
above/below the audible range.

It looks doable with test tones (is this cheating?) but will be more
difficult with music. Would make an interesting little project.