The following survey contains audio clips of English speech with heavy background noise that have been put through two different kinds of Neural Networks in an attempt to isolate the voice. The goal of this survey is to attempt to determine which network produces better results with a fixed amount of training, and also determine the factors that affect their outputs.

For each question, please listen to both audio clips, then answer the proceeding questions. There are 10 pairs in total. You only need to answer the first two questions for each, but we would appreciate it if you answered every question if you have the time.