Abstract-Electroneurographic (ENG) signals were extracted from muscle afferent fibers and used for real-time closed-loop control of FES-based ankle movements in a rabbit preparation. For extraction of the ENG signals, tripolar cuff electrodes were implanted onto the peroneal and tibial nerves in the left hind limb. A neural network was used for extraction of joint angles from the recorded ENGs.For stimulation purposes, percutaneous stainless steel wires were placed intramuscularly into the tibialis anterior and lateral gastrocnemius muscles, respectively. Stimulation intensity was varied by changing the applied pulse width (PW).Step and sinusoidal tracking tasks were performed using a standard PID controller.Results showed that the system's performance is highly sensitive to the initial joint angle; best results were obtained when starting with the ankle joint at a neutral, rest angle. Further, angles estimated from the ENG (by the neural network) lost correlation with measured angles as a given experiment progressed. Improvements were seen when the neural network was allowed to learn intermittently during an experimental session. Finally, a standard PID controller required frequent retuning during an experimental session, which, not surprisingly, suggests that an adaptive controller should be used.Keywords -Natural sensors, neural prosthesis, implanted electrodes, functional electrical stimulation, closed-loop control, artificial neural networks, nerve signals.
I. INTRODUCTIONPersons with paralysis of upper and lower limbs require the use of reliable, robust, closed-loop, and often adaptive devices that would allow them to perform basic tasks by means of applied Functional neuromuscular Electrical Stimulation (FES).However, control of FES-based movements can be enhanced by proper estimation of the current state of motion. For example, in the case of a paralyzed subject walking with the aid of FES, it is crucial that we use reliable sensory information pertaining to the output angular trajectories. The latter information can be obtained from sensory nerve fibers stemming from musculotendinous tissue to the spinal cord (i.e., muscle afferent fibers). A number of features in both time and frequency domains can somehow convey the necessary information albeit in a non-linear and time-variant fashion, and with signal-to-noise ratios much less than unity for cuff electrodes. Thus, our research group has been using implanted cuff electrodes to extract relevant information by analyzing signals obtained directly from the nerve bundles that carry it [1,2,3]. So far, a simple approach has been taken: rectified and bin integrated (RBIN) Electroneurographic (ENG) signals have been monitored and have been found to allow for a reasonable mapping onto angular and torque data by means of neural and fuzzy models [1,2]. The present paper shows our first findings from using the extracted angular information in a closed-loop controller.
II. METHODOLOGY
A. Experimental SetupAcute experiments were conducted with 4 female New Zealand r...