Neuro-Fuzzy Model - Part 2

Abstract

In the previous chapter a feasibility study was presented, which showed the capacity of neuro-fuzzy systems to approximate FES-induced hand movements of stroke subjects. However, the aim of these models was to support the design process of subject-specific forearm surface neuroprostheses, as well as to provide a basis for development of new control techniques. Therefore, the reliability and accuracy of the models should be high and they should adapt to different types of subjects and physiologic characteristics. In order to achieve this, still many aspects of neuro-fuzzy models of FES-induced movements should be tested. In this chapter, the RFNN system that showed the best results in the previous chapter was further analyzed and tested with two different learning methods. The data collected from three stroke subjects and three healthy subjects was used in order to have a heterogeneous subject sample. Data collection was again carried out in collaboration with the ADACEN center.