Towards Generating Diverse Audio Captions via Adversarial Training | IEEE Journals & Magazine | IEEE Xplore