Preliminary results toward a naturally controlled multi-synergistic prosthetic hand

Robotic hands embedding human motor control
principles in their mechanical design are getting increasing
interest thanks to their simplicity and robustness, combined
with good performance. Another key aspect of these hands
is that humans can use them very effectively thanks to the
similarity of their behavior with real hands. Nevertheless,
controlling more than one degree of actuation remains a
challenging task.
In this paper, we take advantage of these characteristics in
a multi-synergistic prosthesis. We propose an integrated setup
composed of Pisa/IIT SoftHand 2 and a control strategy which
simultaneously and proportionally maps the human hand movements
to the robotic hand. The control technique is based on
a combination of non-negative matrix factorization and linear
regression algorithms. It also features a real-time continuous
posture compensation of the electromyographic signals based
on an IMU. The algorithm is tested on five healthy subjects
through an experiment in a virtual environment. In a separate
experiment, the efficacy of the posture compensation strategy is
evaluated on five healthy subjects and, finally, the whole setup
is successfully tested in performing realistic daily life activities.