ROMJIST Volume 21, No. 4, 2018, pp. 377-391
Dan-Marius DOBREA, Monica-Claudia DOBREA A MANFIS model of the 3D head position based on a wearable system
ABSTRACT: The fuzzy systems, the artificial neural networks, and the neuro-fuzzy systems have been widely used in modeling of the complex, unknown and nonlinear functions. In this paper, a comparative study between two soft computing techniques was done, namely, between: (a) a Multiple Neuro-Fuzzy Inference System (MANFIS) and (b) two different Artificial Neural Network structures - and this comparison was one investigated in relation with a practical embedded application. More precisely, the research used a new wearable system, named Intell.TieSens, in order to model the complex nonlinear relationship that exist between the capacitive sensors embedded in a tie collar and the 3D head position of the wearer of that tie. The Intell.TieSens system was designed to be a noncontact system, low power, portable, lightweight, having a Bluetooth Low Energy wireless data communication capability in order to be able both to be paired with a personal computer or with a smartphone and to identify in real time the human 3D head position. The results showed the superior performances obtained with MANFIS system – two times better than the performances obtained with the RBF (Radial Basis Function) neural network. Out of the tested models, the MLP (MultiLayer Perceptron) neural network obtained the lowest performances.KEYWORDS: Fuzzy, MANFIS, CANFIS, neural network, RBF, wearable, capacitive sensingRead full text (pdf)