Geometric algebra and electronics devoted to the control of upper-limb prostheses

The last decades have witnessed significant improvements in the field of prosthetics. However, control strategies are still not as advanced as the prosthetic hardware and fail to provide a robust control to the end user, often leading to device abandonment. This works aspires to improve the control of polyarticulated myoelectric prostheses following two distinct lines of research: a) designing explainable control algorithms for reducing prosthesis recalibration time; b) contributing to the development of a novel implantable EMG-based electronic system. The developed control algorithms are grounded on Geometric Algebra and nearest neighbor classification and are inspired by stroke gesture recognition. These algorithms are the result of the intuition gained from the exploratory work on time series classification and are aimed at reducing the amount of necessary training data and training time while maintaining acceptable recognition rates. It is also intended to provide a geometrically interpretable alternative to the state-of-the-art algorithms, allowing the decision-making process to be more explainable. These algorithms were evaluated on both offline and real-time scenarios and computed high recognition rates (even when compared to other state-of-the-art classifiers) while complying with the reduced training data and training time requirements, indicating that achieving faster device recalibration may be possible. On the other hand, the contributions to the development of the implantable EMG system were based on the development of a firmware/software framework to control different system functionalities and stream/record the EMG signals. Moreover, the feasibility of transcutaneous data transmission using the Bluetooth Low Energy (BLE) protocol was demonstrated, along with a preliminary evaluation of the system prototype. In sum, the two lines of research explored throughout this work share the main aim of providing more intuitive control for prosthetic users, potentially improving their quality of life and device satisfaction.