Harry hissed at her to be quiet.
Harry waved madly at the other three to follow him.
Her and the other three are the semantic Agents of the following infinitives; but I whether they are the grammatical subjects of those infinitives is tricky. What is clearly not the case is this parsing:
✲ [SUBJECT Harry ] [VERB hissed at ] [COMPLEMENT she BE quiet ]
✲ [SUBJECT Harry ] [VERB waved at ] [COMPLEMENT they FOLLOW him ]
That makes no sense. I think rather what we have here are the verbs HISS and WAVE employed bitransitively, taking two complements—an indirect object† representing the person to whom a communication is addressed and a direct object representing the content of the communication. I thus see these sentences as having this structure:
[SUBJECT Harry ] [VERB hissed ] [COMPL-IndObj at her ] [COMPL-DirObj she BE quiet ]
[SUBJECT Harry ] [VERB waved ] [COMPL-IndObj at them ] [COMPL-DirObj they FOLLOW him ]
The final DirObj complement is realized as an infinitival subordinate clause (
she to be quiet, they to follow him) whose subject is deleted because it is the same as the preceding indirect object of the main clause. This is the ‘Equi’ process which Prof. Lawler describes in pellucid detail here.
So the actual subjects of these infinitives have been deleted; but the indirect objects which are left indicate what they would be if they were still there.
✲ marks an utterance as unacceptable
† This will trouble traditional grammarians: the radicals declare that the object of a preposition cannot be an indirect object, while the merely conservative declare that an indirect object must be introduced by to or for. I ground my analysis in analogies like these:*
It's entirely natural for verbs like hiss, wave, bark, growl which have common monotransitive uses with at to carry that preposition over into bitransitive uses.