Machine learning in control of functional electrical stimulation systems for locomotion

Abstract

Two machine learning techniques were evaluated for automatic design of a rule-based control of functional electrical stimulation (FES) for locomotion of spinal cord injured humans. The task was to learn the invariant characteristics of the relationship between sensory information and the FES-control signal by using off-line supervised training. Sensory signals were recorded using pressure sensors installed in the insoles of a subject's shoes and goniometers attached across the joints of the affected leg. The FES-control consisted of pulses corresponding to time intervals when the subject pressed on the manual push-button to deliver the stimulation during FES-assisted ambulation. The machine learning techniques used were the adaptive logic network (ALN) [1] and the inductive learning algorithm (IL) [2]. Results to date suggest that, given the same training data, the IL learned faster than the ALN, while both performed the test rapidly. The generalization was estimated by measuring the test errors and it was better with an ALN, especially if past points were used to reflect the time dimension. Both techniques were able to predict future stimulation events. An advantage of the ALN over the IL was that ALN's can be retrained with new data without losing previously collected knowledge. The advantages of the IL over the ALN were that the IL produces small, explicit, comprehensible trees and that the relative importance of each sensory contribution can be quantified.

Finite state control of functional electrical stimulation for the rehabilitation of gait

“…This, and other studies found that IL identified relationships between sensor values and gait state transitions which were not apparent to a human expert, particularly in the case of anticipatory actions, such as anticipation of the delay between application of FED and generation of the desired mechanical force and movements which were frequently taken by the patient prior to the occurrence of expected gait events (ANDREWS et al, 1992;KOSTOV et al, 1995). Separate studies demonstrated how artificial neural networks (ANNs) could also be used to automatically learn control behaviour from training data KOSTOV et al, 1995).…”

“…Separate studies demonstrated how artificial neural networks (ANNs) could also be used to automatically learn control behaviour from training data KOSTOV et al, 1995). A comparison between IL and ANN methods identified that the IL learning process was relatively quick compared to that of adaptive logic networks, a type of ANN , and resulted in systems which were generally simpler than their ANN counterparts .…”

Electrical Stimulation and Motor Recovery

“…Sacral nerve stimulation (SNS) has been used since 1967 to activate bladder voiding (91,132,182,241), to treat bladder incontinence (20,34,44,59,78,108,119,120,122,221,229), to alleviate pelvic pain (1,38,190,279), to prevent fecal incontinence (50,(72)(73)(74)103,153), and to relieve constipation (56,128,133,134,161,166,167,186,189,248,252,272 (87,112,138,147,198,199,219) led to many devices (152,201) for improving walking (81,237), activating hand function (24) through voice commands (94), enhancing coughing (154), and even relieving seating pressures (65). Called neuroprostheses (25,193), these devices often are designed to correct specific deficits, such as foot drop (27,217,245) and hand grasp weakness (47,249,267).…”

External Sensors for Detecting the Activation and Deactivation Times of the Major Muscles Used in Walking

“…For the purpose of event detection, sensors should provide a signal that can be used to turn stimulation of different muscles on and off accurately and reliably. Ideally, this detection should use a simple threshold crossing technique to limit the computational load required, although neural nets [24], Kalman filtering [17], and other techniques may also work in real time. Finally, sensors should function well over a range of speeds.…”