This paper presents a lightweight neural architecture for efficient direction-of-arrival (DoA) estimation, built upon a novel Ghost Squeeze-and-Excitation Residual Block (GSERB). The proposed block integrates Squeeze-and-Excitation (SE) attention directly into the Ghost feature generation process, enabling selective channel enhancement while minimizing redundant feature computation. Unlike conventional attentionaugmented convolutional networks, the GSERB amplifies informative ghost feature channels prior to fusion with intrinsic representations, thereby improving feature discrimination and representational efficiency. Experimental results showcase that this integration enhances the model's ability to capture spa-tial-spectral dependencies while maintaining a compact design. The resulting network, named GSERB-DoANet, achieves an effective balance between accuracy, scalability, and computational efficiency for real-time multi-source DoA estimation in wireless sensing applications.

This paper explores the application of Temporal Fusion Transformers (TFTs) to target recognition and multi-horizon tracking, with a particular emphasis on enhancing Direction of Arrival (DoA) estimation in complex signal environments. By leveraging the advanced temporal modeling capabilities and dynamic feature selection mechanisms inherent to the TFT architecture, we adapt and optimize the model for improved performance in antenna signal processing, leading to a more accurate and robust estimation framework. Our analysis underscores the architectural strengths of TFTs, especially their capacity to model long-range temporal dependencies and to dynamically prioritize sensor-derived features—capabilities that are essential for precise target tracking and recognition. We train and evaluate the model under diverse conditions, including multi-horizon forecasting and varying prediction latencies across models. The proposed method is benchmarked against state-of-the-art algorithms, including recurrent neural networks and standard transformer-based models. Extensive experimental results demonstrate that the TFT-based approach consistently outperforms existing techniques in forecasting accuracy, multi-target recognition, and computational complexity, highlighting its potential to advance the field of DoA estimation.

As virtual reality (VR) and 360° video applications continue to gain traction, multi-quality tile-based 360° video streaming has emerged as a flexible and adaptive solution for delivering high-quality immersive content while optimizing bandwidth usage. However, the presence of mixed quality levels within a single 360° video frame can adversely affect users' perceived quality. In this work, we investigate the impact of spatial quality variations across tiles and within users' viewports, and provide improved perceptual quality modeling for tiled 360° videos. To realize this goal, we introduce the Subjective Tile based Assessment for 360° Videos (STAV360) dataset, which includes six source 360° videos, each encoded using twelve tile based configurations, subjective ratings from 27 participants, and corresponding viewing trajectories. Based on this dataset, we propose novel tile-based and viewport-aware 360° video quality assessment (360° VQA) models that incorporate tile-level quality information and user interaction data. Experimental results show that our models outperform conventional full-frame methods by more accurately predicting perceived quality. This work provides key insights for designing perceptually-driven quality metrics and lays the foundation for more efficient and adaptive tile-based 360° video streaming strategies.

The implementation of indoor localization is being facilitated through commercially available low-cost mm-wave sensors. These sensors generate point cloud outputs containing noisy estimates of detected targets due to hardware noise and multipath reflections. In contrast to previously studied regression approaches, this study introduces a classification-based approach to predict the angle-of-arrival (AoA) and range of a human target from point clouds obtained from an mm-wave sensor. Our proposed methodology achieves a 7% and 26% improvement in AoA and range prediction, respectively, compared to the baseline models. All experiments have been conducted and validated using real data recorded by the mm-wave sensor.

The autonomous navigation of UAS requires, among others, detect-and-avoid capability as a prerequisite for enabling wide-ranging applications, including the transportation of goods and people. This article presents the design, implementation, and experimental results of a UAS-borne radar system detecting drones. The applications of the proposed system include not only detect-and-avoid systems for safe and autonomous navigation of unmanned aircraft systems but also airborne surveillance of malicious drones in controlled or restricted airspace for mitigating security and privacy threats. The system performance in terms of maximum detection range is evaluated through field tests. The experimental results show that the proposed UAS-borne radar can detect a DJI Phantom 4 and a DJI Matrice 600 Pro at a maximum distance of 440 and 500 meters, respectively. The article also provides insights into the system implementation and integration aspects, discusses future research direction, and stresses the need for standardization efforts to benchmark the required performance levels for UAS-borne radars.