Low-power radar-based system for materials and objects recognition

Conventional prosthetic technology for upper limb amputations still suffers from major limitations. Semi-autonomous (i.e., self-grasping) prosthetic hands have been proposed as a promising solution to mitigate some of those limitations. In an effort to contribute to the development of semi-autonomous prosthetic hands, this work explores radars and artificial intelligence algorithms for object recognition. In particular, Pulse-coherent radars and Deep Learning are explored here as potential approaches for in-hand object recognition. The Integration of three radars was undertaken, with a specific emphasis on sensors' timing synchronization to prevent interferences. Subsequently, a detailed data collection and analysis was conducted to recognize the selected items. Finally, offline and real-time investigation into the recognition performance was carried out showing promising results. These findings could be of value for the integration of such low-power, low-cost sensors into prosthetic hands for embedded, real-time object recognition and subsequent autonomous grasping facilitating daily use for individuals with upper limb loss.