Question: Do you want results from specific mixtures?
Answer: We have given many options for people to really have a chance to do well. As a minimum you should do 2x2 unmixing filter matrix. Try as many rooms as you can. If you can do 4x4 unmixing filters that would be even better. No need to do over-complete.
Question: May we change the sampling rate?
Answer: Yes, just let us know how you did it so we can reproduce your result.
Question: How silent were your 'silent' speakers when you recorded the data?
Answer: We did not turn the 'silent' speakers off, but we believe that the speakers were sufficiently silent. At least, we could not recognize any hum, but perhaps a little hiss.
Question: Where do you intend to publish your review paper?
Answer: See call for papers.
Question: Have any groups published any results on the same data?
Answer: This is new data. No results have been published on this yet.
Question: Are you restricting solutions to time-invariant linear filters?
Answer: Time-invariant should be enough as the recording setup is stationary. We do not intend to record moving sources.
Question: Are you giving any thought or attention as to hardware software complexity?
Answer: In the review we should discuss this but for the evaluation everything goes.
Question: How will the approval process work exactly? What information will you give back to those that submit solutions about the solutions provided by others?
Answer: We will provide all numbers that are available without the identity of the teams that submitted. If you then agree to disseminate your results (or part of them) they will be public to all.
Question: What is to keep someone from using LMS to filter the sole source recordings from room 1 out of the mixed recordings, and supplying those filters?
Answer: We intend to use LMS using the test data. Anyone can do this with the information provided for room 1. This room is provided as an exercise. We felt there should be at least one room on which anyone can evaluate the performance of their algorithm.