Audio Set: An ontology and human-labeled dataset for audio events

  Abstract