Centre for Digital Music

 
 

IEEE AASP Challenge:
Detection and Classification of Acoustic Scenes and Events

Results

D-CASE Overall Results Poster (PDF)
Scene Classification Results
Event Detection (Office Live) Results
Event Detection (Office Synthetic) Results

Data and code

Publicly available challenge datasets

Source code for some of the submitted systems

Source code for the baseline systems: scene recognition; event detection

Introduction

We invite researchers in signal processing, machine learning and other fields to participate in our challenge, which consists of a set of related tasks on automatic detection and classification of acoustic scenes and acoustic events.

The tasks fall into the field of computational auditory scene analysis (CASA). Humans are able to follow specific sound sources in a complex audio environment with ease and the development of systems that try to mimic this behaviour is an open problem, especially in the case of overlapping sound events.

The deadline for confirming participation to the challenge is 31st March 2013 (please send an email to aasp-challenge-owner@eecs.qmul.ac.uk with participant names/affiliations). The deadline for code submission is 14th April 2013.

Results will be presented at a special session in WASPAA 2013; participants are invited to present a poster at a special session. Also, authors of novel work are encouraged to submit their work as a regular paper at WASPAA 2013.

In order to participate in the discussion phase and to receive news about the challenge, please subscribe to the aasp-challenge mailing list.

Submission Instructions

Basic Steps

  • Be sure to read the technical report explaining the challenge motivation, datasets, and metrics
  • Be sure to read the specifications for the tasks you will be participating (links below)
  • Develop your system(s) using the publicly available challenge datasets. You can use the metrics functions to test your system's performance.
  • Write an extended abstract outlining what your submission is trying to accomplish (links below)
  • Verify that your code works using the AASP Challenge Linux System Image (for non-Matlab submissions)
  • Confirm your participation to the challenge via email by 31st March
  • Submit your code and extended abstract using either email or via a download link by 14th April
  • Present a poster at the challenge poster session at WASPAA 2013

Task Specifications

Scene Classification Specifications (pdf)
Event Detection (Office Live) Specifications (pdf)
Event Detection (Office Synthetic) Specifications (pdf)

Extended Abstract Details

The extended abstracts provide the outside world with a general understanding of what each submission is trying to accomplish. The extended abstracts need NOT be cutting edge/never-before-published materials. Participants are encouraged to submit novel work as a regular paper at the WASPAA 2013 conference.

The extended abstracts can be revised by the authors after the results have been announced. We will post the final versions of the extended abstracts in the challenge website. The challenge abstracts:

  • Should be 2-3 pages long
  • Must conform to the guidelines in the following templates: LaTeX template   Word template
  • Must be submitted in PDF format.
  • Can include references to other publications about your work (if existing)

Organisers

This challenge is organised by the Centre for Digital Music and by IRCAM, under the auspices of the Audio and Acoustic Signal Processing (AASP) technical committee of the IEEE Signal Processing Society. Click here for the original challenge proposal document.

The organisers,
Dimitrios Giannoulis (QMUL), Emmanouil Benetos (City University London), Dan Stowell (QMUL), Mathias Rossignol (IRCAM), Mathieu Lagrange (IRCAM) and Mark D. Plumbley (QMUL)

More on AASP challenges

SPS_logo