This paper proposes a new approach for wide angle monitoring of vital signs in smart home applications. The person is tracked using an indoor radar. Upon detecting the person to be static, the radar automatically focuses its beam on that location, and subsequently breathing and heart rates are extracted from the reflected signals using continuous wavelet transform ($CWT$) analysis. In this way, leveraging the radar’s on-chip processor enables real-time monitoring of vital signs across varying angles. In our experiment, we employ a commercial multi-input multi-output (MIMO) millimeter-wave FMCW radar to monitor vital signs within a range of $1.15$ to $2.3$ m and an angular span of $-44.8$ to $+44.8$ deg. In the Bland–Altman plot, the measured results indicate the average difference of $-1.5$ and $0.06$ beats per minute (BPM) relative to the reference for heart rate and breathing rate, respectively.

Fall incidents represent a major public health problem among elderly people. This resulted in a significant increase of the number of investigated systems aiming at detecting falls promptly. In this respect, in this work, a biomedical radar system is proposed for remote real-time fall detection and indoor localization. The system, consisting of a sensor and a base station, combines radar and wireless communication techniques, and uses a data processing technique to distinguish between fall events and normal movements. The classification, based on a Least-Square Support Vector Machine (LS -SVM), combined with the sliding window principle allows to perform fall detection in real-time. Moreover, it is capable to localize the subjects when the fall incident has been detected. The in-vivo validation showed a high success rate in detecting fall events, with a maximum delay of 340 ms. Moreover, a maximum mean absolute errors (MAE) of 3.8 cm and a maximum root-mean-square error (RMSE) of 7.5 cm were reported in measuring the subject's absolute distance.

Indoor radar-based human activity recognition (HAR) using machine learning has shown promising results. However, deploying an HAR model in unseen environments remains challenging due to a potential mismatch between training and operational conditions. Such mismatch can be reduced by acquiring annotated training data in more diverse situations. However, since this is time intensive, this paper explores the application of data augmentation and unsupervised domain adaptation (UDA) to enhance the robustness of HAR models, even when they are trained using a very limited amount of annotated data. In the initial analysis, a baseline HAR model was evaluated using a validation set (a) from the same environment as the training data and (b) from a different environment. The results showed a 29.6% decrease in the F1-score when tested on data from the different environment. Implementing data augmentation techniques—specifically, time–frequency warping—reduced this performance gap to 17.8%. Further improvements were achieved by applying an unsupervised domain adaptation strategy, which brought the performance gap drop down to 13.2%. Furthermore, an ablation study examining various augmentation methods and synthetic sample quantities demonstrates the superior performance of our proposed augmentation approach. The paper concludes with a discussion on how environmental variations, such as changes in aspect angle, occlusion and layout, can affect the time-Doppler radar representation and, consequently, HAR performance.

This article introduces an innovative approach integrating machine learning (ML) methods into a far-field wireless power transfer (WPT) system. Within this system, a radar is employed to detect the presence and height of individuals. These data, combined with information on the location of nodes, as well as the propagation channel between transmitters and nodes, are fed into the proposed ML algorithm. The ML model aims to predict the optimal power level for each transmitter, effectively establishing a safe 3-D zone around people while also maximizing power transfer efficiency. The advantages of using ML are the realization of a real-time system, which is crucial in indoor applications, keeping dangerous radiation at a safe level with a very low risk of harmful exposure, and simultaneously enhancing efficiency. Three ML models are evaluated, namely the random forest (RF), support vector machine (SVM), and neural network (NN). Simulation results highlight the superior performance of the NN model, demonstrating its ability to effectively capture the complex nonlinear characteristics of indoor propagation environments, with only approximately 6% of its predictions exceeding the predefined safety threshold. The experimental results show that NN-based WPT can maintain the electric field amplitude (EFA) below a defined threshold for multiple indoor experimental scenarios over the person’s height. In addition, the proposed approach outperformed the maximum ratio transmission (MRT) approach in terms of radio frequency–radio frequency (RF–RF) transmission efficiency in 21.43% of the measurements conducted with multiple people present in the testbed.

Ensuring the safety of electromagnetic exposure stands as an important concern in wireless power transfer (WPT) systems. This work proposes a distributed Fusion Radar WPT (FRWPT) system designed to maintain safe Electric Field Amplitude (EFA) levels at specific locations detected by the radar, primarily where an individual is present. This approach allows for higher EFA in areas without the person, thus optimizing overall power utilization within the system. Also, the radar's ability to detect a person's velocity allows for projecting the person's upcoming location to ensure safety in advance. We introduce an algorithm including power weighting factors for controlling power to not only mitigate dangerous radiation but also maximize power utilization. One significant challenge is the estimation of EFA considering multipath propagation, a common issue in indoor environments. To overcome this, we explore the indoor EFA distribution and suggest a simulation-based method for EFA estimation, taking into account the amplifying effect of the human body on EFA. Experimental results demonstrate that the system successfully maintains EFA below a predefined threshold across various human locations. Moreover, these experiments highlight the system's capability to maximize power utilization ratio (PUR), achieving a value exceeding 50%.

Prospective authors are requested to submit new, unpublished manuscripts for inclusion in the upcoming event described in this call for papers.