Feasibility of Using Neuro-Fuzzy Subject-Specific Models for Functional Electrical Stimulation Induced Hand Movements

Abstract

Functional Electrical Stimulation (FES) is a technique that artificially elicits muscle contractions and it is used to restore motor/sensory functions in both assistive and therapeutic applications. The use of multi-field surface electrodes is a novel popular approach in transcutaneous FES applications. Lately, hybrid systems that combine artificial neural networks and fuzzy logic have also been proposed for many applications in different areas. This paper presents the possibility of combining both approaches for obtaining subject-specific models of FES induced hand movements for grasping applications. Data of the hand and finger motion from two subjects affected by acquired brain injury were used to train two different approaches: coactive neuro-fuzzy inference system and recurrent fuzzy neural network. Preliminary results show that these approaches can be considered in modelling applications for their ability to learn and predict main characteristics of the system, as well as providing useful information from the original system that could be interpreted as subject-specific knowledge.