Thank you for taking part in the listening test.
You are asked to evaluate the performance of source separation algorithms that attempt to separate the vocal from a mixture. The ideal result would be if the algorithm could perfectly extract the Reference vocals, so you are asked to rate how similar each test item is to this idea. The Original mix is also included to help guide your judgement, but it is the Reference vocals that you are asked to compare with, not the Original mix. In other words, the highest-rated stimulus should sound as close as possible to the Reference vocals track only.
The “next” button will take you to the tutorial and familiarisation pages, where you can listen to various samples of the stimuli, and then to the actual test pages, where you are asked to grade the stimuli based on the overall similarity to the original vocal sound.
Please feel free to pause at any time if you have any question or need any assistance in taking the test.