This paper delves into the transformative potential of Machine Learning (ML) in Radio Frequency (RF) sensing applications. We focus on pivotal domains such as device localization, occupancy detection, activity monitoring, and biometric sensing, showcasing how ML is redefining the boundaries of what is possible. By harnessing the power of ML, we showcase how to unlock unprecedented performance enhancements in these critical areas. We provide a comprehensive review of cutting-edge ML-driven RF sensing methodologies and offer an overview of publicly available datasets that are propelling this field forward. Moreover, we present key challenges that remain - from the quality and labeling of RF sensor data to robustness, privacy, and explainability of ML models. Through this exploration, we lay the path for future scientific and engineering innovations in the ever-evolving landscape of RF sensing.

In this paper, we present an ultra low-power Wake-up Receiver front-end operating in the 868/915 MHz ISM band. It targets short distance body area networks. Its power consumption is only 126 uW, including a low-power on-chip ring oscillator. Since the receiver targets small transmission distances, up to 10 m, sensitivity is traded against power consumption. This is achieved by removing the LNA and making all the gain at the low IF frequencies. The receiver sensitivity is -65 dBm at a BER of 0.1%.

In this work, we propose a signal processing technique for beam-steering radar architectures allowing concurrent two-dimensional (2-D) localization and vital signs monitoring of human subjects. We demonstrated it by using a single-input single-output (SISO) frequency-modulated continuous wave (FMCW) radar which integrates two frequency-scanning antennas (FSAs). This method is capable of isolating the Doppler signal generated by each single subject from the contributions of all the reflections in the monitored environment. This allows determining the number of individuals in the room and accurately measuring their vital signs parameters (respiration and heart rates) and 2-D positions (range and azimuth information). The spectral analysis, the data matrix generation and the signal processing technique are detailed and discussed. Experimental results demonstrated the feasibility of the proposed approach, showing the ability in determining the number of subjects present in the room, in accurately measuring and tracking over time their vital signs parameters, and in 2-D localization with errors within the limits of the radar range and angular resolutions. Practical applications arise for healthcare, Hospital 4.0, Internet of Medical Things (IoMT), ambient assisted living, smart buildings and through-wall sensing.

A near-field galvanic coupled transdural telemetry ASICs for intracortical brain-computer interfaces is presented. The proposed design features a two channels transmitter and three channels receiver (2TX-3RX) topology, which introduces spatial diversity to effectively mitigate misalignments (both lateral and rotational) between the brain and the skull and recovers the path loss by 13 dB when the RX is in the worst-case blind spot. This spatial diversity also allows the presented telemetry to support the spatial division multiplexing required for a high-capacity multi-implant distributed network. It achieves a signal-to-interference ratio of 12 dB, even with the adjacent interference node placed only 8 mm away from the desired link. While consuming only 0.33 mW for each channel, the presented RX achieves a wide bandwidth of 360 MHz and a low input referred noise of 13.21 nV/√ H z. The presented telemetry achieves a 270 Mbps data rate with a BER<10 ^-6 and an energy efficiency of 3.4 pJ/b and 3.7 pJ/b, respectively. The core footprint of the TX and RX modules is only 100 and 52 mm2, respectively, minimizing the invasiveness of the surgery. The proposed transdural telemetry system has been characterized ex-vivo with a 7-mm thick porcine tissue.

In this paper we present an ultra-low-power receiver geared towards body area networks (BAN). The presented wideband-FSK receiver consumes only 382.5μW while achieving a BER of 10 ^-3 at -81dBm sensitivity for 12.5kbps. The bit rate is scalable up to 625kbps, enabling a trade-off between sensitivity and bit rate. Taking advantage of the short-range nature of BAN applications, a mixer-first architecture is proposed, leading to a good dynamic range, given the DC power consumption. To further decrease the power consumption a free-running digitally controlled oscillator (DCO), tunable from 782MHz to 932MHz, is implemented, that is controlled by a data-aided automatic frequency control (AFC) loop, making the receiver resilient against DCO frequency variations.

The efficiency of an energy receiver of a Wireless Power Transfer (WPT) system is intricately influenced by many factors such as frequency, power, topology, and the parameters of the rectifying diode. Developing an efficient and accurate model is critical for effective design. Unlike analytical models that rely on signal approximation and circuit simplification, the time-domain model can comprehensively incorporate all diode parameters and provide a complete view of the circuit with all time-domain signals available. However, to ensure accuracy and rapid convergence, the output low-pass section has to be handled carefully due to its DC voltage bias for the diodes and its long time constant. Furthermore, scaling the time-domain model for diverse rectenna topologies presents challenges. In this work, we propose a recursive time-domain model suitable for half-wave and any-stage voltage multiplier rectennas. The output low pass section is addressed by replacing it with an equivalent DC voltage source, thus avoiding the long settling time and ensuring accurate DC bias for the diodes. An analytical model for converting input voltage to DC calculates the equivalent source. Our proposed model is scalable for any number of multiplier stages and can accommodate the effect of matching networks. We validate the accuracy of calculated power conversion efficiency and rectifier input impedance by comparing them with results from Harmonic Balance simulations.