eduzhai > Applied Sciences > Engineering >

An Acoustic Segment Model Based Segment Unit Selection Approach to Acoustic Scene Classification with Partial Utterances

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: In this paper, we propose a sub-utterance unit selection framework to removeacoustic segments in audio recordings that carry little information foracoustic scene classification (ASC). Our approach is built upon a universal setof acoustic segment units covering the overall acoustic scene space. First,those units are modeled with acoustic segment models (ASMs) used to tokenizeacoustic scene utterances into sequences of acoustic segment units. Next,paralleling the idea of stop words in information retrieval, stop ASMs areautomatically detected. Finally, acoustic segments associated with the stopASMs are blocked, because of their low indexing power in retrieval of mostacoustic scenes. In contrast to building scene models with whole utterances,the ASM-removed sub-utterances, i.e., acoustic utterances without stop acousticsegments, are then used as inputs to the AlexNet-L back-end for finalclassification. On the DCASE 2018 dataset, scene classification accuracyincreases from 68 , with whole utterances, to 72.1 , with segment selection.This represents a competitive accuracy without any data augmentation, and orensemble strategy. Moreover, our approach compares favourably to AlexNet-L withattention.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×