Review
Chemistry, Multidisciplinary
Konstantinos Papadopoulos, Mohieddine Jelali
Summary: The importance of radar-based human activity recognition has increased significantly due to its advantages in vision-based sensing in poor environmental conditions, along with increased public sensitivity to privacy protection and progress in cost-effective manufacturing. This review outlines the basics and recent advances in both classical machine learning and deep learning-based human activity recognition, and evaluates their performance and computational effort based on benchmarking dataset.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Electrical & Electronic
Sahil Waqar, Muhammad Muaaz, Matthias Patzold
Summary: Modern monostatic radar-based HAR systems perform well in detecting human activities towards or away from the radar, but fail to classify multidirectional activities. In this article, a distributed MIMO radar configuration is proposed to overcome this limitation by capturing and analyzing multidirectional human movements from multiple perspectives.
IEEE SENSORS JOURNAL
(2023)
Article
Computer Science, Theory & Methods
Fuqiang Gu, Mu-Huan Chung, Mark Chignell, Shahrokh Valaee, Baoding Zhou, Xue Liu
Summary: This study provides a comprehensive survey on recent advances and challenges in human activity recognition (HAR) with deep learning, highlighting the lack of in-depth research on deep learning methods in HAR.
ACM COMPUTING SURVEYS
(2021)
Article
Engineering, Electrical & Electronic
Ziyu Liu, Chaoyang Wu, Wenbin Ye
Summary: This article proposes a radar-based human activity recognition (HAR) model that achieves high classification accuracy through few-shot learning. By extracting feature vectors using a trained feature extractor and utilizing the classification weights of the most similar original category, the classification weight of the new human activity category is generated quickly. Additionally, a cosine similarity-based classifier is employed to address the issue of weight value interval mismatch, and a large margin softmax cross-entropy (LMSC) loss function is used to enhance the model's performance.
IEEE SENSORS JOURNAL
(2022)
Article
Computer Science, Theory & Methods
Kaixuan Chen, Dalin Zhang, Lina Yao, Bin Guo, Zhiwen Yu, Yunhao Liu
Summary: This study explores the impact of sensor devices and the Internet of Things on sensor-based activity recognition, discusses the role of deep learning methods in addressing recognition challenges, and provides an overview of current research progress and future directions.
ACM COMPUTING SURVEYS
(2021)
Article
Environmental Sciences
Mingyue Lu, Yuchen Li, Manzhu Yu, Qian Zhang, Yadong Zhang, Bin Liu, Menglong Wang
Summary: Accurate and timely precipitation forecasts are important for decision-making, planning, and protecting lives and property. This research proposes a multisource ConvLSTM (MS-ConvLSTM) model that incorporates multiple data sources, including composite reflectivity, echo top, vertically integrated liquid water, and radar-retrieved wind field data, to improve the accuracy of precipitation forecasting. Experimental results showed that the proposed model outperformed traditional methods in terms of various evaluation metrics.
Article
Environmental Sciences
Yanlin Li, Freddy Galindo, Julio Urbina, Qihou Zhou, Tai-Yin Huang
Summary: We present MADAME, a machine-learning approach for detecting and analyzing meteor echoes, using advanced machine-learning techniques including supervised and unsupervised learning. Our results show that YOLOv4, a CNN-based object detection model, performs well in detecting and identifying meteor head and trail echoes. MADAME can autonomously process data in interferometer mode and determine the target's radiant source and vector velocity.
Article
Engineering, Electrical & Electronic
Zhongping Cao, Zhenchang Li, Xuemei Guo, Guoli Wang
Summary: This paper addresses the issue of adapting radar-based human activity recognition systems to new environments without source data. By utilizing the source hypothesis transfer learning architecture, a mechanism for cross-environment adaptation of radar-based HAR is developed. A reliable self-supervised labeling strategy is proposed to generate pseudo labels for unlabeled target data, leading to the improvement of target-specific feature extraction for environment adaptation. The experimental results demonstrate the effectiveness of the proposed approach on a public HAR dataset.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2021)
Article
Automation & Control Systems
Emanuele Lattanzi, Chiara Contoli, Valerio Freschi
Summary: Deep learning is widely considered as the cutting-edge technology in many fields, but its high accuracy comes at the cost of demanding computational resources. To address this issue, early exit, a design methodology, is proposed to trade off accuracy for latency, reducing the burden on resources. This study explores the application of early exit in human activity recognition tasks and evaluates its impact on different deep network architectures and communication technologies. The experimental results show that early exit significantly improves latency (up to 35x) without sacrificing accuracy in most cases, confirming its viability as an adaptive approach. However, in distributed environments, the benefits of early exit vary depending on the model's inference latency and the extent of the use of far exit points.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Geochemistry & Geophysics
Simin Zhu, Ronny Gerhard Guendel, Alexander Yarovoy, Francesco Fioranelli
Summary: In this paper, a method for unconstrained human activities recognition with a radar network is proposed. It combines convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to extract spatial-temporal patterns. Different data fusion methods are explored and compared for utilizing the rich information provided by the radar nodes. The experimental results show that the proposed classifier achieves high accuracy in nine-class human activity recognition.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Information Systems
Tatsuhito Hasegawa, Kazuma Kondo
Summary: Sensor-based human activity recognition (HAR) is an important technology in IoT services. HAR using representation learning is widely used due to the difficulty of extracting meaningful features from raw sensor data. This study proposes an easy ensemble (EE) for HAR, which allows deep ensemble learning in a single model. Various techniques (input variationer, stepwise ensemble, and channel shuffle) for the EE are also introduced. Experiments on a benchmark data set demonstrate the effectiveness of EE and its techniques compared with conventional ensemble learning methods.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Computer Science, Artificial Intelligence
Ioannis Vernikos, Evaggelos Spyrou, Ioannis-Aris Kostis, Eirini Mathe, Phivos Mylonas
Summary: In real-life scenarios, occlusion of body parts in human activity recognition from video data is often underestimated. This work presents a new approach for recognizing human activities under partial occlusion, considering up to two occluded body parts. It formulates the problem as a regression task and uses a novel deep Convolutional Recurrent Neural Network (CRNN) to reconstruct the missing information. Experimental results demonstrate a significant performance improvement compared to baseline approaches. This is the first research work that addresses the problem of HAR under occlusion as a regression task.
INTERNATIONAL JOURNAL OF NEURAL SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Yujia Qian, Chuan Chen, Longzhen Tang, Yong Jia, Guolong Cui
Summary: In this article, a recognition method based on a multispectrogram and deep-learning model is proposed for human activity recognition (HAR) based on radar. The method transforms radar echo data into spectrograms using three time-frequency analyses and utilizes a parallel LSTM-CNN network (PLCN) to learn both temporal and spatial features of the spectrograms, achieving a high recognition accuracy of 94.75% for eight human activities.
IEEE SENSORS JOURNAL
(2023)
Article
Geochemistry & Geophysics
Xinyu Li, Yuan He, Francesco Fioranelli, Xiaojun Jing
Summary: The paper introduces a semi-supervised transfer learning algorithm JDS-TL for radar-based HAR, which successfully alleviates the need for labeling a large number of radar signals by using a sparsely labeled dataset. Experiments show that JDS-TL achieves an average accuracy of 87.6% in recognizing six activities with only 10% labeled instances, highlighting the efficiency of domain adaptation and semantic transfer modules.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Lei Jin, Xiaojuan Wang, Jiaming Chu, Mingshu He
Summary: Recently, Human Activity Recognition (HAR) based on sensors using deep learning networks has gained popularity due to its wide application. However, the inter-personal variability in human activity data poses challenges to both closed-set classification and open-set problem. In this study, we propose a framework that incorporates a loss function of Euclidean distance and a high-dimensional embedding layer to extract discriminative features. We also define two types of open-set problems and propose corresponding solutions. Experimental results demonstrate that our method significantly improves the performance of the models.
IEEE SENSORS JOURNAL
(2022)
Article
Telecommunications
Yuan He, Wanqing Huang, Haoyan Wei, Hongtao Zhang
Summary: This letter analyzes the impact of channel fading on handover performance in UAV networks by introducing a path-loss-plus-fading model with handover parameters. By discretizing the handover states during time-to-trigger duration and modeling fading as Nakagami-m distribution, handover failure and ping-pong probabilities are derived through analyzing handover state probabilities. The results show a tradeoff between handover failure and ping-pong probabilities when configuring the time-to-trigger duration.
IEEE COMMUNICATIONS LETTERS
(2021)
Correction
Telecommunications
Yuan He, Wanqing Huang, Haoyan Wei, Hongtao Zhang
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Telecommunications
Jianghui Liu, Hongtao Zhang, Yuan He
Summary: This letter proposes a dynamic tunable model for adjusting the serving radius of unmanned aerial vehicle base stations, along with a semi-progressive offloading deployment scheme to optimize UAV number and overlapping interference. By covering a certain proportion of ground terminals first and then adjusting the serving radius, the scheme reduces UAV quantity and interference effectively. Fine adjustments of power or antenna angles are utilized to expand the coverage area and reduce overlapping interference.
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Geochemistry & Geophysics
Xinyu Li, Yuan He, Francesco Fioranelli, Xiaojun Jing, Alexander Yarovoy, Yang Yang
Summary: The article introduces an instance-based transfer learning method (ITL) to address the issue of limited radar data samples by utilizing limited radar micro-Doppler (MD) features. Experimental results demonstrate that ITL performs well with limited training samples, outperforming other transfer learning methods, and has better generalization performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Geochemistry & Geophysics
Xinyu Li, Yuan He, Francesco Fioranelli, Xiaojun Jing
Summary: The paper introduces a semi-supervised transfer learning algorithm JDS-TL for radar-based HAR, which successfully alleviates the need for labeling a large number of radar signals by using a sparsely labeled dataset. Experiments show that JDS-TL achieves an average accuracy of 87.6% in recognizing six activities with only 10% labeled instances, highlighting the efficiency of domain adaptation and semantic transfer modules.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Wenfei Tang, Hongtao Zhang, Yuan He
Summary: This paper proposes a method for interference coordination using 3D blockage effects in urban environments, which organizes a dynamic UAV group to serve users and implements interference control within the group, thereby improving network coverage performance.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2022)
Article
Chemistry, Analytical
Hongdi Liu, Hongtao Zhang, Yuan He, Yong Sun
Summary: This paper proposes a two-level jamming decision-making framework based on dual Q-learning model to optimize the jamming strategy and dynamic evaluation method, achieving improvement in radar jamming effectiveness by learning joint strategy of mode switching and parameter agility.
Article
Engineering, Electrical & Electronic
Yun Jing Zhang, Ying Tian, Mei Song Tong, Yuan He
Summary: This paper presents a new sensing method based on passive generalized Parity-Time (GPT)-symmetry, which enhances sensitivity compared to traditional methods in the microwave frequency range. The method utilizes a GPT-symmetry circuit to establish a sensor, and investigates the perturbation of equivalent lumped elements to reflection coefficients, showing improved sensitivity in terms of resonant frequency shift and reflection magnitude change compared to conventional methods. The operating frequency can also be tuned by the coupling coefficient. An experiment is conducted to measure glucose water concentrations, verifying the method's sensitivity in terms of resonant frequency and reflection magnitude change. The results demonstrate good consistency with simulations and highlight the advantages of long-distance noncontact measurement with high sensitivity.
IEEE SENSORS JOURNAL
(2022)
Article
Engineering, Electrical & Electronic
Wenfei Tang, Hongtao Zhang, Yuan He, Mingyu Zhou
Summary: This paper proposes a 3D coordination model for interference management in multi-antenna UAV networks, using multi-cell beamforming and signal-level cooperation. The system's performance is analyzed by deriving a semi-closed expression of coverage probability using stochastic geometry. The results provide optimal deployment parameters for UAV deployments and show that the coverage probability can reach 92% when SIR threshold T = 0 dB.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
(2022)
Article
Engineering, Electrical & Electronic
Xinyu Li, J. Andrew Zhang, Kai Wu, Yuanhao Cui, Xiaojun Jing
Summary: This article proposes three algorithms for Doppler frequency estimation based on the ratio of channel state information. These algorithms explore different properties of the CSI ratio and accurately estimate the Doppler frequency in bistatic setups with clock asynchronism.
IEEE SENSORS JOURNAL
(2022)
Article
Engineering, Electrical & Electronic
Yuan He, Li Zhang, Mei Song Tong
Summary: This study focuses on the high-resolution inner imaging of objects that possess both dielectric and magnetic properties, such as mineral substances, which is important for detecting and analyzing their ingredients. An integral equation method is developed using microwave illumination to reconstruct or image these objects. Full volume integral equations (VIEs) are employed to describe the problem since the objects are both dielectric and magnetic. The objects are reconstructed by alternatively solving the forward scattering VIEs (FSVIEs) and the inverse scattering VIEs (ISVIEs) using the Born iterative method (BIM) or distorted BIM (DBIM). The Nystrom method is used to solve the FSVIEs, and the Gauss-Newton minimization method (GNMM) with a multiplicative regularization scheme (MRS) is used to solve the ISVIEs.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION
(2023)
Article
Engineering, Electrical & Electronic
Heng Yang, Shi Cong Wang, Peng Li, Yuan He, Yun Jing Zhang
Summary: In this study, a reflective polarization conversion metasurface (PCM) with symmetrical L-shaped patches is proposed, enabling switch between linear-to-linear (LTL) cross-polarization and linear-to-circular (LTC) polarization modes using PIN diodes. The PCM demonstrates ultrawide bandwidth for both LTL and LTC polarization conversion through multimode activation. A microstrip line-based dc biasing network composed of radial stubs is introduced to isolate RF currents, overcoming issues caused by RF-lumped choke inductors in the ultrawide bandwidth. A prototype with 15 x 15 units is fabricated and shows successful LTC and LTL polarization conversion within specific frequency ranges.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION
(2023)
Article
Engineering, Electrical & Electronic
Yun Jing Zhang, Jing Lei Yong, Ying Tian, Yuan He
Summary: This article presents a new measurement method for liquid concentrations using two inductively coupled resonators, which has higher sensitivity compared to traditional methods. The sensitivity of the center eigenfrequency shift and the reflection magnitude change to the concentration can reach 9.4 MHz/(mg/mL) and 5.41 dB/(mg/mL), respectively. A sensor composed of an open quarter-wavelength microstrip-line resonator (OQMR) and a split-ring resonator (SRR) is designed for detecting water-glucose solution concentrations.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Geochemistry & Geophysics
Jianping Wang, Runlong Li, Yuan He, Yang Yang
Summary: In this article, a prior-guided deep learning approach is proposed for interference mitigation in FMCW radars. A complex-valued convolutional neural network is utilized, and a prior feature is exploited as a regularization term to improve performance and convergence.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)