"The problem I see with this is that they implant monkey before paralysis and can use machine learning to build a model of the monkey's unique representation for walking. Then they paralyze and show the monkey's can adapt to only using their interface in 5 days. There's no way the mapping from neural activity to stimulation is generalizable..." [1] [1] https://www.reddit.com/r/science/comments/5c2ljz/brain_impla...
But maybe it is. You can also put people on a VR environment and have them "walk" there (see experiments by Nicolelis et. al.) and try to capture signals like that