I guess the problem lies in detection of impulse from brain. The device is capable of full range of movement... But there's missing signal to commit movement as such.
Further I can guess is taking 2D furrier and then make a furrier for changes in frequencies specific's amplitudes, and than maybe again.
With some recurrsion and then feed it maybe as 1D array with formatting with spaces, kept correlations between each other into some livenet, that would recognizes intended movement, and control it by headband...
I guess nowadays it's somewhere connected to leftover neurons, which lacks certain imaginability as they correspond to motor center for muscles, and that got to be translated to machine control.
But I think it's today's tech.
