eduzhai > Applied Sciences > Engineering >

A Sequential Self Teaching Approach for Improving Generalization in Sound Event Recognition

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 14 pages

Abstract: An important problem in machine auditory perception is to recognize anddetect sound events. In this paper, we propose a sequential self-teachingapproach to learning sounds. Our main proposition is that it is harder to learnsounds in adverse situations such as from weakly labeled and or noisy labeleddata, and in these situations a single stage of learning is not sufficient. Ourproposal is a sequential stage-wise learning process that improvesgeneralization capabilities of a given modeling system. We justify this methodvia technical results and on Audioset, the largest sound events dataset, oursequential learning approach can lead to up to 9 improvement in performance. Acomprehensive evaluation also shows that the method leads to improvedtransferability of knowledge from previously trained models, thereby leading toimproved generalization capabilities on transfer learning tasks.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...