SLOW-FAST AUDITORY STREAMS FOR AUDIO RECOGNITION

Evangellos Kazakos; Arsha Nagrani; Andrew Zisserman; Dima  Damen

SLOW-FAST AUDITORY STREAMS FOR AUDIO RECOGNITION

Evangellos Kazakos, Arsha Nagrani, Andrew Zisserman, Dima Damen

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

110 Downloads (Pure)

Abstract

We propose a two-stream convolutional network for audio recognition, that operates on time-frequency spectrogram inputs. Following similar success in visual recognition, we learn Slow-Fast auditory streams with separable convolutions and multi-level lateral connections. The Slow pathway has high channel capacity while the Fast pathway operates at a fine-grained temporal resolution. We showcase the importance of our two-stream proposal on two diverse datasets: VGG-Sound and EPIC-KITCHENS-100, and achieve state- of-the-art results on both.

Original language	English
Title of host publication	ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Number of pages	5
ISBN (Electronic)	978-1-7281-7605-5
ISBN (Print)	978-1-7281-7606-2
Publication status	E-pub ahead of print - 13 May 2021

Publication series

Name
ISSN (Print)	2379-190X

Keywords

training
visualization
time-frequency analysis
convolution
channel capacity
conferences
speech recognition
audio recognition
action recognition
fusion
multi-stream networks

Access to Document

Full-text PDF (author’s accepted manuscript)
This is the author accepted manuscript (AAM). The final published version (version of record) is available online via IEEE at https://ieeexplore.ieee.org/document/9413376 . Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 968 KB

https://ieeexplore.ieee.org/document/9413376Licence: Unspecified

Handle.net

Persistent link

UMPIRE: United Model for the Perception of Interactions for visual Recognition
Damen, D. (Principal Investigator)
1/02/20 → 31/01/25
Project: Research

Outstanding Paper - ICASSP 2021
Damen, D. (Recipient) & Kazakos, V. (Recipient), Jun 2021
Prize: Prizes, Medals, Awards and Grants

HPC (High Performance Computing) and HTC (High Throughput Computing) Facilities
Alam, S. R. (Manager), Williams, D. A. G. (Manager), Eccleston, P. E. (Manager) & Greene, D. (Manager)
Facility/equipment: Facility

Cite this

@inproceedings{62912bc6bdd94d5195245082c6d87556,

title = "SLOW-FAST AUDITORY STREAMS FOR AUDIO RECOGNITION",

abstract = "We propose a two-stream convolutional network for audio recognition, that operates on time-frequency spectrogram inputs. Following similar success in visual recognition, we learn Slow-Fast auditory streams with separable convolutions and multi-level lateral connections. The Slow pathway has high channel capacity while the Fast pathway operates at a fine-grained temporal resolution. We showcase the importance of our two-stream proposal on two diverse datasets: VGG-Sound and EPIC-KITCHENS-100, and achieve state- of-the-art results on both.",

keywords = "training, visualization, time-frequency analysis, convolution, channel capacity, conferences, speech recognition, audio recognition, action recognition, fusion, multi-stream networks",

author = "Evangellos Kazakos and Arsha Nagrani and Andrew Zisserman and Dima Damen",

year = "2021",

month = may,

day = "13",

language = "English",

isbn = "978-1-7281-7606-2",

publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

booktitle = "ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)",

address = "United States",

}

SLOW-FAST AUDITORY STREAMS FOR AUDIO RECOGNITION. / Kazakos, Evangellos; Nagrani, Arsha; Zisserman, Andrew et al.
ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Institute of Electrical and Electronics Engineers (IEEE), 2021.

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - SLOW-FAST AUDITORY STREAMS FOR AUDIO RECOGNITION

AU - Kazakos, Evangellos

AU - Nagrani, Arsha

AU - Zisserman, Andrew

AU - Damen, Dima

PY - 2021/5/13

Y1 - 2021/5/13

N2 - We propose a two-stream convolutional network for audio recognition, that operates on time-frequency spectrogram inputs. Following similar success in visual recognition, we learn Slow-Fast auditory streams with separable convolutions and multi-level lateral connections. The Slow pathway has high channel capacity while the Fast pathway operates at a fine-grained temporal resolution. We showcase the importance of our two-stream proposal on two diverse datasets: VGG-Sound and EPIC-KITCHENS-100, and achieve state- of-the-art results on both.

AB - We propose a two-stream convolutional network for audio recognition, that operates on time-frequency spectrogram inputs. Following similar success in visual recognition, we learn Slow-Fast auditory streams with separable convolutions and multi-level lateral connections. The Slow pathway has high channel capacity while the Fast pathway operates at a fine-grained temporal resolution. We showcase the importance of our two-stream proposal on two diverse datasets: VGG-Sound and EPIC-KITCHENS-100, and achieve state- of-the-art results on both.

KW - training

KW - visualization

KW - time-frequency analysis

KW - convolution

KW - channel capacity

KW - conferences

KW - speech recognition

KW - audio recognition

KW - action recognition

KW - fusion

KW - multi-stream networks

M3 - Conference Contribution (Conference Proceeding)

SN - 978-1-7281-7606-2

BT - ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

PB - Institute of Electrical and Electronics Engineers (IEEE)

ER -

SLOW-FAST AUDITORY STREAMS FOR AUDIO RECOGNITION

Abstract

Publication series

Keywords

Access to Document

Handle.net

Fingerprint

Projects

UMPIRE: United Model for the Perception of Interactions for visual Recognition

Prizes

Outstanding Paper - ICASSP 2021

Equipment

HPC (High Performance Computing) and HTC (High Throughput Computing) Facilities

Cite this