• Initial Conditions
• Inverse problem
• Levenberg-Marquardt
• Magnetic Modeling
• Magnetic Tracking
• Numerical Optimization
• PSO
• Swarm Intelligence

In this work, we focus on a blind localization strategy for the myokinetic interface, that is a novel paradigm for the design of a bi-directional human-machine interface (HMI) for prosthetic applications. Core of this interface is the use of permanent magnets either to decode the intentions of the user and to induce phantom illusions to guarantee sensory feedback. Previous studies on this interface proved its feasibility and mainly investigate on the localization of multiple magnetic markers, providing also a first embedded prototype of the localizer. However, most magnetic tracking systems (included the one implemented for this interface) adopt fast iterative algorithms which suffer from the choice of the initial conditions. Indeed, these algorithms are commonly fed with the initial reference positions and orientations of the magnets that allow the retrieval of their position for the first time. This thesis provides the development of a magnetic tracking algorithm able to retrieve the position of the magnets without any knowledge of their positioning (just limiting the search on a sensing volume within the sensor workspace). Furthermore, this thesis provides a detailed study on the choice of magnetic models for the development of a generic localizer, with particular attention to the constraints of the myokinetic interface.