Guest Editorial: Advances in AI-assisted radar sensing applications

Abstract

Recent developments in Artificial Intelligence (AI) and the accessibility of cost-effective radar hardware have transformed various sectors, including e-healthcare, smart cities, and critical infrastructures. AI holds immense potential for enhancing radar technology. However, there are significant challenges hindering its adoption in this domain. These challenges encompass Radar Data Accessibility, which involves limited access to radar data for training AI models due to low sample availability. Data Labelling, requiring domain-specific expertise, and Data Pre-processing, aimed at selecting the best radar data representation for AI applications, are complex and vital steps. Additionally, integrating an AI framework into radar hardware, whether using pre-trained or custom models, presents a major obstacle. This special issue focuses on research, articles, and experiments that bridge the gap between radar hardware and AI frameworks, addressing these critical challenges.

The special issue has garnered significant interest, with a total of 13 paper submissions. After rigorous peer review, nine papers that met high publication standards were accepted. These papers collectively address crucial challenges in AI-assisted radar technology, offering innovative ideas, insightful analyses, and experimental results that bridge the gap between radar hardware and AI frameworks. Most notably, these papers include real-world validation and demonstrate innovative system designs and processing solutions. They advance current knowledge and pave the way for future innovations in the field.

Among the featured papers, Zhou et al. focus on the application of millimetre-wave radar, specifically 4D TDM MIMO FMCW radar, for health monitoring and human activity recognition [1]. Their comprehensive simulation model achieves an impressive 90% average classification accuracy, offering valuable insights for radar configuration and activity testing. Zhenghui Li et al. introduce an innovative approach to radar-based human activity recognition across six domains, with adaptive thresholding and holistic optimisation, significantly improving classification accuracy [2]. Li et al. propose a ground-breaking voice identification method using Ultra-Wideband technology, leveraging micro-Doppler shifts during speech production to achieve close to 90% accuracy in healthcare applications [3].

Yu et al. explore radar-based human activity recognition for elderly care health monitoring, addressing noisy radar signals. They introduce wavelet denoising and the Double Phase Cascaded Denoising and Classification Network, improving accuracy and robust activity monitoring [4]. Xiong et al. tackle track-to-track association (T2TA) challenges by using homography estimation to address radar bias, enhancing association credibility and reducing manual labelling [5]. Perďoch et al. utilise a simple Neural Network (NN) for signal pre-processing, effectively reducing clutter and enhancing processing speed in radar systems [6].

Yang et al. discuss the use of millimetre-wave radar technology to alleviate traffic congestion, achieving a 20% reduction through radar sensors and advanced traffic state discrimination algorithms [7]. Rehman et al. contribute to millimetre-wave radar systems for marine autonomy by distinguishing maritime targets from sea clutter using experimental data from field trials [8]. Wu et al. develop a hybrid NN model for target recognition against reflector jamming, outperforming other methods in recognition performance and robustness [9].

These papers collectively advance the field of radar technology and signal processing, offering practical solutions to real-world challenges.