Decoding force from multiunit recordings from the median nerve

Much attention has been focused on the detection of volitionary motor commands from the efferent Peripheral Nervous System as a control signal for an advanced prosthetic limb, or the delivery of artificial sensory data to the Peripheral Nervous System as feedback. Less explored has been the potential for natural sensory signals to act as sensor input to neuroprosthetic systems. Many conditions with paralysis as a symptom leave the afferent peripheral nervous system functional, and potentially available as a feedback signal to a control system. In order to demonstrate the feasibility of using such a signal we decode a multiunit afferent nerve signal and use an extreme learning machine to perform a regression to decode force data. From this we were able to show that afferent signals from the fingertip can be decoded into force profiles.