The 2013 Signal Separation Evaluation Campaign
Nobutaka Ono, Zbynek Koldovsky, Shigeki Miyabe, Nobutaka Ito

Abstract:
This paper summarizes the 2013 community-based Signal Separation Evaluation Campaign (SiSEC 2013). Five speech and music datasets were contributed, including two new datasets: ``Two-channel noisy recordings of a moving speaker within a limited area" and ``Asynchronous recordings of speech mixtures". The participants addressed one or more datasets out of five, and the results for each task were evaluated using different objective performance criteria. We overview the campaign specifications such as the tasks, datasets, and evaluation criteria. We also summarize the evaluated performance of the submitted systems, and discuss organization strategies for future campaigns.