We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Sound

Title: Acoustic Features Fusion using Attentive Multi-channel Deep Architecture

Abstract: In this paper, we present a novel deep fusion architecture for audio classification tasks. The multi-channel model presented is formed using deep convolution layers where different acoustic features are passed through each channel. To enable dissemination of information across the channels, we introduce attention feature maps that aid in the alignment of frames. The output of each channel is merged using interaction parameters that non-linearly aggregate the representative features. Finally, we evaluate the performance of the proposed architecture on three benchmark datasets:- DCASE-2016 and LITIS Rouen (acoustic scene recognition), and CHiME-Home (tagging). Our experimental results suggest that the architecture presented outperforms the standard baselines and achieves outstanding performance on the task of acoustic scene recognition and audio tagging.
Comments: Accepted in CHiME'18 (Interspeech Workshop)
Subjects: Sound (cs.SD); Audio and Speech Processing (eess.AS)
Cite as: arXiv:1811.00936 [cs.SD]
  (or arXiv:1811.00936v1 [cs.SD] for this version)

Submission history

From: Gaurav Bhatt [view email]
[v1] Fri, 2 Nov 2018 15:31:22 GMT (281kb,D)

Link back to: arXiv, form interface, contact.