Automatic people counting has garnered significant attention due to its broad civilian and military applications. In civilian settings, it helps detect unusual occupancy patterns or manage crowding in public transportation. In military contexts, it serves to count and track enemy movements, providing real-time data on troop numbers and positions on the battlefield, which is critical for tactical decision-making. Radar systems are often used for such tasks due to their ability to function in all weather conditions, day or night. However, the signal collected by the radar is hindered by unwanted signals reflected by clutter. Also, the direct coupling between transmit and receive antennas can mask targets with a weak signal. All these artifacts can decay the performance of deep learning models for automatic people counting. This work proposes a background mitigation algorithm based on the multiresolution analysis of the maximal overlap discrete wavelet transform (MRA-MODWT) to enhance the accuracy of deep learning models for automatic people counting. Subsequently, the Daubechies least asymmetric wavelet with four vanishing moments (sym4) is used to isolate and cancel background signals, and a fusion layer combining a transfer learning block with a customized deep convolutional neural network (DCNN) is introduced to improve the accuracy. The hybrid DCNN-InceptionV3 model achieved a peak accuracy of 98.31%, an average precision of 0.9827, an average recall of 0.9827, and an average <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula> score of 0.9836 on realistic radar data.

Human vocal signals are essential for information exchange. Recently, research has been conducted on capturing vocal signals not only through microphones but also using radar. While mode decomposition methods are a representative approach for enhancing radar-based vocal signals, their performance is often compromised by a critical dependency on manually-selected parameters. This paper proposes a novel framework, composite mode fitness score-successive variational mode decomposition (CMFS-SVMD), to overcome this limitation. We utilize a 77 GHz frequency-modulated continuous-wave (FMCW) multiple-input multiple-output (MIMO) radar and Curve Length (CL) method for human localization. The core of our work is the CMFS-SVMD which adaptively and automatically selects the optimal balancing parameter for SVMD by minimizing a novel fitness score tailored to vocal signal characteristics. The performance of the proposed algorithm is validated by comparing the extracted fundamental frequency F<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> against a ground truth derived from a synchronized microphone, using the root mean square error (RMSE) as the primary metric. Experimental results demonstrate that our proposed algorithm accurately tracks the F<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> of various utterances, including words, sentences, and sustained vowels, proving its robustness and adaptability.

This letter addresses the problem of beam pattern re-synthesis due to Transmit/Receive Module (TRM) malfunctions in active phased array antenna systems. Since TRM failures in full array systems cause beam pattern distortion and degrade system performance, a re-synthesis technique to compensate for these failures is essential. Conventional optimization algorithms, such as the Genetic Algorithm (GA) and Particle Swarm Optimization (PSO), suffer from high computational costs that hinder real-time application, while deep learning-based approaches are limited by the requirement for large-scale training datasets. To resolve these issues, this letter proposes a transfer learning-based beam pattern re-synthesis technique. The proposed method pre-trains the physical principles of beamforming using the array factor and derives compensation weights by fine-tuning with a small amount of optimization data. In addition, a compound loss function is designed to prevent gain degradation during the re-synthesis process. Simulation results demonstrate that the proposed technique achieves a 40% lower Mean Squared Error (MSE) compared to a model without pre-training using only 1,000 fine-tuning samples, while recovering the peak gain to 97.9% of the nominal level.

When an electronic device featuring an enclosure and an internal PCB is located inside a building, analyzing the coupling phenomenon resulting from external high-power electromagnetic waves at specific locations on the PCB using conventional commercial simulation tools requires significant computational resources. As the analysis frequency increases beyond several GHz, the number of mesh elements grows substantially, making full electromagnetic simulation practically infeasible. This paper proposes a spatially partitioned frequency-domain analysis method based on the Huygens’ box-based electromagnetic topology approach to overcome the limitations of conducting coupling analysis for small enclosures inside buildings. First, a Huygens’ box is assumed to contain the enclosure of the electronic equipment, and the coupling characteristics of the external electromagnetic waves incident through a window into the Huygens’ box are analyzed. Then, a separate Huygens’ box and enclosure are constructed to analyze the coupling effects at specific locations on the internal PCB. Additionally, a two-stage Huygens’ box is introduced to contain the PCB inside the enclosure, and the results are compared. Finally, the feasibility and validity of the proposed partitioned-recombination analysis method are verified by comparing its results with those of full-model simulations up to 10 GHz.

An A6-type relativistic magnetron (RM) with a conventional diffraction output (DO) structure generates the TE31 mode that is unsuitable for antenna applications because of its poor central directivity and degraded beam focusing efficiency. To address this limitation, the TE31 mode must be converted into the fundamental TE11 mode that is known for its high directivity. In this study, we designed and comparatively analyzed three types of TE31 to TE11 mode converters using CST Microwave Studio: two converters employing a length difference method (one converter with a perpendicular and the other converter with a parallel waveguide arrangement) and one converter using a coupled slot structure. The simulation results demonstrated that the coupled slot-based converter achieved the highest conversion efficiency of 97 % and a wide bandwidth. Furthermore, among the two length difference designs, the parallel waveguide configuration exhibited a superior conversion efficiency of 90 % compared with 86 % for the perpendicular configuration.

In this paper, we describe a ray-tracing method for spherical lenses and reestablish an equation for calculating the radiation pattern using a spherical lens based on previous literature by modifying the physical optics (PO) integral equation. Using this formula, we calculated the radiation pattern according to the size and point source distance of a single-medium spherical lens and a Luneburg lens. While both lenses showed that the beam width became narrower as the lens size increased, the side-lobe level decreased when the source was located on the surface of the lens in the case of the Luneburg lens and when the source was appropriately distanced from the lens surface in the case of the single-medium spherical lens. In addition, through the analysis of the focal length and the beam forming the focus when parallel rays are incident on a single-medium spherical lens, we show that a single-medium spherical lens with a refractive index of 1.4 to 1.5 can replace a complex Luneburg lens by placing the source at a distance of 0.4 λ to 0.5 λ from the lens surface.

Recently, frequency-modulated continuous wave (FMCW) multiple-input multiple-output (MIMO) radar has been employed indoors for various applications, including target localization, automated people counting, and human vital signs detection. However, the signals collected by this radar technology are significantly affected by artifacts such as clutter and noise. The primary factors degrading radar performance include the direct coupling between transmit and receive antennas and the presence of stationary clutter within the experimental environment, collectively referred to as the background effect in radar imagery. This work aims to propose a space-time approach to mitigate the background effect. Specifically, the background is characterized in the frequency-wavenumber (f-k) domain using a mask function that attenuates magnitude spectrum components not corresponding to the background. The modified magnitude spectrum is then converted back to the space-time domain for background suppression. To evaluate the performance of the proposed method, realistic measurements were conducted in two scenarios involving human targets in continuous back-and-forth motion within a room. The results demonstrate effective background reduction and improved robustness to target signal degradation compared to other space-time approaches such as moving average, single-delay moving target indication (MTI), and high-pass filtering methods.

Automatic people counting has garnered significant attention due to its broad civilian and military applications. In civilian settings, it helps detect unusual occupancy patterns or manage crowding in public transportation. In military contexts, it serves to count and track enemy movements, providing real-time data on troop numbers and positions on the battlefield, which is critical for tactical decision-making. Radar systems are often used for such tasks due to their ability to function in all weather conditions, day or night. However, the signal collected by the radar is hindered by unwanted signals reflected by clutter. Also, the direct coupling between transmit and receive antennas can mask targets with a weak signal. All these artifacts can decay the performance of deep learning models for automatic people counting. This work proposes a background mitigation algorithm based on the multiresolution analysis of the maximal overlap discrete wavelet transform (MRA-MODWT) to enhance the accuracy of deep learning models for automatic people counting. Subsequently, the Daubechies least asymmetric wavelet with four vanishing moments (sym4) is used to isolate and cancel background signals, and a fusion layer combining a transfer learning block with a customized deep convolutional neural network (DCNN) is introduced to improve the accuracy. The hybrid DCNN-InceptionV3 model achieved a peak accuracy of 98.31%, an average precision of 0.9827, an average recall of 0.9827, and an average <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula> score of 0.9836 on realistic radar data.