• embedded systems
• DSP
• space
• assistive technology
• deep learning
• artificial intelligence
• edge computing
• FPGA
• COTS
• market
• robotic arm
• KWS
• cloud detection

Nowadays, the conflicting necessities to exploit emerging technologies, such as Deep Learning, and to process data on the edge embedded platforms, featuring limited computing resources, has exacerbated the complexity of design trade-offs in Digital Signal Processing (DSP) design. Moreover, additional constraints are set by the peculiarities of niche market applications, whose reduced sales volumes and intrinsic requirements often prevent R&D prototypes from improving their technology maturity. Because of that, in this work, we investigate the design of DSP-based systems with respect to several case studies in the field of space and assistive technology.
In the field of assistive technology, our research questions involve the implementation of a portable Voice User Interface (VUI) for Italian speakers with severe dysarthria and the design of a low-effort Human-Machine Interface to control a power wheelchair-mounted manipulator to assist people with upper-limb disabilities.
In the field of space applications, we faced the problem of the limited downlink throughput of Earth Observation (EO) satellites through the deployment of Deep Neural Networks (DNNs) on board space-craft to compress their downlink data.
As a side project, we showed a flexible implementation of the CCSDS 131-2-B-1 standard, a payload telemetry transmitter ensuring high downlink throughput thanks to the use of adaptive coding and modulation.