In this report we describe the creation of a random, data-blind system to provide a random baseline for Task 3 (sound event detection in real life audio) in the DCASE 2016 challenge. Particular attention is paid to the results of two sound events occurring in the residential area scene, one very rare, the other very frequent. The relatively good performance of the random system in comparison to the results of the proper detection systems shows the difficulty of Task 3 given the current state-of-the-art sound detection methods.

Report (PDF)

  ⤧  Next post Huang et al. (2017) ⤧  Previous post Computer learns to recognize sounds by watching video