Towards reconstructing intelligible speech from the human auditory cortex (nature.com)

Is this pre- or post-processing? Are we reading what the ears hear or what the brain hears? If we can actually hear what the brain hears, is it possible that that could lead to better noise-reduction algorithms?