Slashdot Mirror
Amputees Control Virtual Prosthetic Arm Using Nerve Signals (newscientist.com)
CanadianRealist writes: Current prosthetic arms are usually controlled by detecting signals from the user twitching muscles in the shoulder or arm. This allows only a limited number of possible movements, such as grasp and release. Researchers have developed a new technique that interprets signals from motor neurons in the spinal cord, allowing for a greater range of control of an arm. Signals from nerves associated with hand and arm movements were mapped to the corresponding movements. Test subjects were able to move a virtual prosthetic arm with greater freedom than has been achieved with muscle-controlled prosthetics. (Note: A virtual prosthetic arm was used rather than a real one as this work is still in the early stages.) The study has been published in the journal Nature Biomedical Engineering.
https://www.youtube.com/watch?...
https://www.youtube.com/watch?...
Liberty - Security - Laziness - Pick any two.
This is a very different approach to interpreting the EMG than traditional myoelectric prostheses. They're using targeted muscle reinnervation with multi-electrode arrays to decompose the EMG into individual motor units, then using the more precise motor unit signals to control. ME control couples a preserved, non-involved muscle to control of a single prosthesis motor, where targeted reinnervation gives them access to neural signal intended for the missing muscles. Decomposing a complex EMG into its component motorneurons works well with fine wire electrodes, where the recording volume is fairly small, but surface EMG contains too much information/noise. Using the MEA to decipher that complex signal is pretty clever.
It looks like it only worked well in 7/9 cases, and it's not clear whether they repeated over different days and different electrode placements. It's a proof-of-concept study showing that you can decompose surface EMG into a higher fidelity signal than just patterns of intensity