Chendong Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Chendong Xu has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Qisong Wu
Southeast University
6 shared publications - Shuai Yao
Key Laboratory of Underwater Acoustic Signal Processing of Ministry of Education, Southeast University, Nanjing 210096, China
5 shared publications - Yong Wang
Key Laboratory of Underwater Acoustic Signal Processing of the Ministry of Education, Southeast University, Nanjing, China
5 shared publications - Dongyu Liu
Key Laboratory of Underwater Acoustic Signal Processing of the Ministry of Education, Southeast University, Nanjing, China
4 shared publications - Kuiying Yin
Nanjing Research Institute of Electronics Technology, Nanjing, China
2 shared publications Bao Zhang
Key Laboratory of Underwater Acoustic Signal Processing (Southeast University), Ministry of Education, Nanjing, China
1 shared publication
Latest Publications
- 26 June 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (26) , 266111
Single-channel respiratory signal-based respiratory event detection using a dual-stream multi-scale encoder network
Sleep apnea/hypopnea syndrome (SAHS) is a widespread disorder associated with serious health risks. Although polysomnography is the diagnostic gold standard, its use is limited by high costs, extended duration, and labor-intensive procedures. To provide a more efficient solution, we propose DSENet, a novel dual-stream encoder-based network that enables automatic SAHS detection using only single-channel respiratory signals. DSENet employs a hierarchical partitioning strategy, dividing input signals into patches and units to effectively capture both coarse-grained and fine-grained respiratory patterns. Furthermore, the integration of the multi-scale feature aggregation module facilitates effective extraction of multi-scale global and local features, while an adversarial decoupling classifier increases robustness across different populations. Extensive experiments demonstrate that DSENet not only efficiently models complex temporal dependencies, but also consistently outperforms existing methods. In particular, using the nasal airflow signal, DSENet achieved 86.29% accuracy and 76.74% F1-score on Multi-Ethnic Study of Atherosclerosis, and 98.18% accuracy and 97.79% F1-score on Apnea-ECG. Code is publicly available at https://github.com/eecdxu/Respiratory-Event-Detection .
- 26 June 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (26) , 266111
Single-channel respiratory signal-based respiratory event detection using a dual-stream multi-scale encoder network
Sleep apnea/hypopnea syndrome (SAHS) is a widespread disorder associated with serious health risks. Although polysomnography is the diagnostic gold standard, its use is limited by high costs, extended duration, and labor-intensive procedures. To provide a more efficient solution, we propose DSENet, a novel dual-stream encoder-based network that enables automatic SAHS detection using only single-channel respiratory signals. DSENet employs a hierarchical partitioning strategy, dividing input signals into patches and units to effectively capture both coarse-grained and fine-grained respiratory patterns. Furthermore, the integration of the multi-scale feature aggregation module facilitates effective extraction of multi-scale global and local features, while an adversarial decoupling classifier increases robustness across different populations. Extensive experiments demonstrate that DSENet not only efficiently models complex temporal dependencies, but also consistently outperforms existing methods. In particular, using the nasal airflow signal, DSENet achieved 86.29% accuracy and 76.74% F1-score on Multi-Ethnic Study of Atherosclerosis, and 98.18% accuracy and 97.79% F1-score on Apnea-ECG. Code is publicly available at https://github.com/eecdxu/Respiratory-Event-Detection .
- 14 March 2026
- Published by Elsevier in Signal Processing
- Vol. 245
A joint sparse Bayesian approach to multi-range-bin DoA estimation for integrated sensing and communication system
- 14 March 2026
- Published by Elsevier in Signal Processing
- Vol. 245
A joint sparse Bayesian approach to multi-range-bin DoA estimation for integrated sensing and communication system
- 2 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-15
Fine-Grained Contactless Human Respiratory Measurement Using Millimeter-Wave Radar
Respiration is a critical vital sign, and its continuous, fine-grained monitoring provides essential insights for clinical assessment and daily health management. Millimeter-wave radar offers a promising contactless and privacy-preserving solution for respiratory monitoring. However, accurately reconstructing respiratory waveforms from radar echoes remains challenging due to the distributed scattering of the human thorax resulting in the respiratory signals across multiple range bins, especially in a high range-resolution radar system, as well as interferences from the body micro-motions such as random body movements (RBMs). In this paper, we propose mmBreath, a fine-grained respiration monitoring system based on a 60 GHz frequency-modulated continuous wave (FMCW) radar. mmBreath adaptively integrates phase signals from multiple range bins to capture comprehensive respiratory dynamics, and employs an attention-enhanced variational autoencoder (VAE) to reconstruct high-fidelity respiratory waveforms. Furthermore, we introduce a respiratory spectrogram reconstruction task as an auxiliary objective to incorporate time-frequency features, which effectively suppresses interference from RBMs, thus further improving reconstruction accuracy. Extensive experiments on 17 hours of data from 12 human subjects demonstrate that mmBreath achieves superior accuracy and robustness, highlighting its potential for practical, contactless respiratory monitoring in real-world scenarios.
- 2 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-15
Fine-Grained Contactless Human Respiratory Measurement Using Millimeter-Wave Radar
Respiration is a critical vital sign, and its continuous, fine-grained monitoring provides essential insights for clinical assessment and daily health management. Millimeter-wave radar offers a promising contactless and privacy-preserving solution for respiratory monitoring. However, accurately reconstructing respiratory waveforms from radar echoes remains challenging due to the distributed scattering of the human thorax resulting in the respiratory signals across multiple range bins, especially in a high range-resolution radar system, as well as interferences from the body micro-motions such as random body movements (RBMs). In this paper, we propose mmBreath, a fine-grained respiration monitoring system based on a 60 GHz frequency-modulated continuous wave (FMCW) radar. mmBreath adaptively integrates phase signals from multiple range bins to capture comprehensive respiratory dynamics, and employs an attention-enhanced variational autoencoder (VAE) to reconstruct high-fidelity respiratory waveforms. Furthermore, we introduce a respiratory spectrogram reconstruction task as an auxiliary objective to incorporate time-frequency features, which effectively suppresses interference from RBMs, thus further improving reconstruction accuracy. Extensive experiments on 17 hours of data from 12 human subjects demonstrate that mmBreath achieves superior accuracy and robustness, highlighting its potential for practical, contactless respiratory monitoring in real-world scenarios.
- 18 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 3695-3699
GAWNet: A Gated Attention Wavelet Network for Respiratory Monitoring via Millimeter-Wave Radar
Millimeter-wave radar has attracted increasing attention for respiratory monitoring due to its non-contact operation and privacy-preserving characteristics. Nevertheless, extracting fine-grained respiratory waveforms from non-stationary radar signals remains highly challenging, as these signals are frequently contaminated by various interferences, most notably aperiodic body micromotion. The spectral components of such interference often overlap with the respiratory frequency band and typically exhibit power levels that significantly exceed the target signal. This letter introduces the Gated Attention Wavelet Network (GAWNet), an interpretable framework that integrates deep learning with physical priors by operating on radar phase information in the wavelet domain. GAWNet leverages a two-stage suppression strategy: first, a Temporal Gated Attention (TGA) encoder combines convolutional gating and self-attention to achieve initial interference reduction; second, a Frequency Gated Attention (FGA) decoder provides further refinement by transforming wavelet coefficients to the frequency domain for precise filtering. The clean respiratory waveform is then reconstructed using an Inverse Discrete Wavelet Transform (IDWT). Extensive experiments with data from 12 subjects demonstrate that GAWNet consistently outperforms state-of-the-art models and exhibits robust generalization capability.
- 18 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 3695-3699
GAWNet: A Gated Attention Wavelet Network for Respiratory Monitoring via Millimeter-Wave Radar
Millimeter-wave radar has attracted increasing attention for respiratory monitoring due to its non-contact operation and privacy-preserving characteristics. Nevertheless, extracting fine-grained respiratory waveforms from non-stationary radar signals remains highly challenging, as these signals are frequently contaminated by various interferences, most notably aperiodic body micromotion. The spectral components of such interference often overlap with the respiratory frequency band and typically exhibit power levels that significantly exceed the target signal. This letter introduces the Gated Attention Wavelet Network (GAWNet), an interpretable framework that integrates deep learning with physical priors by operating on radar phase information in the wavelet domain. GAWNet leverages a two-stage suppression strategy: first, a Temporal Gated Attention (TGA) encoder combines convolutional gating and self-attention to achieve initial interference reduction; second, a Frequency Gated Attention (FGA) decoder provides further refinement by transforming wavelet coefficients to the frequency domain for precise filtering. The clean respiratory waveform is then reconstructed using an Inverse Discrete Wavelet Transform (IDWT). Extensive experiments with data from 12 subjects demonstrate that GAWNet consistently outperforms state-of-the-art models and exhibits robust generalization capability.
- 15 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (16) , 31290-31301
Hierarchical Transformer With Auxiliary Learning for Subject-Independent Respiration Emotion Recognition
Respiration is modulated by human emotional activity. Emotion recognition using physiological signals has recently gained considerable attention. However, most existing studies primarily focus on using electroencephalogram (EEG) signals for emotion recognition. This paper explores the potential of utilizing respiration signals collected by wearable devices for emotion recognition. We propose a hierarchical Transformer model to effectively extract emotional information from respiration signals. Furthermore, we introduce gender classification as an auxiliary task to further improve the accuracy of emotion recognition. Specifically, a Frame Transformer is employed to capture emotional information across frames of respiration signals. The extracted frame-level features are subsequently fused with segment-level embeddings through a specially designed fusion layer. Next, separate Segment Transformers are employed for emotion and gender to extract segment-level information, with a cosine similarity loss applied to promote shared feature learning. Finally, distinct classifiers are used for emotion and gender classification. To evaluate the effectiveness of the proposed method, extensive experiments are conducted on the DEAP and MAHNOB-HCI datasets under a subject-independent setting. For the DEAP dataset, the average classification accuracies are 72.42% for valence and 73.91% for arousal, while for the MAHNOB-HCI dataset, they are 80.45% and 79.69%, respectively. Compared to other physiological signals such as EEG, respiration signals exhibit comparable potential for emotion recognition.
- 15 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (16) , 31290-31301
Hierarchical Transformer With Auxiliary Learning for Subject-Independent Respiration Emotion Recognition
Respiration is modulated by human emotional activity. Emotion recognition using physiological signals has recently gained considerable attention. However, most existing studies primarily focus on using electroencephalogram (EEG) signals for emotion recognition. This paper explores the potential of utilizing respiration signals collected by wearable devices for emotion recognition. We propose a hierarchical Transformer model to effectively extract emotional information from respiration signals. Furthermore, we introduce gender classification as an auxiliary task to further improve the accuracy of emotion recognition. Specifically, a Frame Transformer is employed to capture emotional information across frames of respiration signals. The extracted frame-level features are subsequently fused with segment-level embeddings through a specially designed fusion layer. Next, separate Segment Transformers are employed for emotion and gender to extract segment-level information, with a cosine similarity loss applied to promote shared feature learning. Finally, distinct classifiers are used for emotion and gender classification. To evaluate the effectiveness of the proposed method, extensive experiments are conducted on the DEAP and MAHNOB-HCI datasets under a subject-independent setting. For the DEAP dataset, the average classification accuracies are 72.42% for valence and 73.91% for arousal, while for the MAHNOB-HCI dataset, they are 80.45% and 79.69%, respectively. Compared to other physiological signals such as EEG, respiration signals exhibit comparable potential for emotion recognition.
- Open Access
- 17 November 2022
- Published by MDPI AG in Applied Sciences
- Vol. 12 (22) , 11684
High-Resolution Through-the-Wall Radar Imaging with Exploitation of Target Structure
It is quite challenging for through-the-wall radar imaging (TWRI) to achieve high-resolution ghost-free imaging with limited measurements in an indoor multipath scenario. In this paper, a novel high-resolution TWRI algorithm with the exploitation of the target clustered structure in a hierarchical Bayesian framework is proposed. More specifically, an extended spike-and-slab clustered prior is imposed to statistically encourage the cluster formations in both downrange and crossrange domains of the target region, and a generative model of the proposed approach is provided. Then, a Markov Chain Monte Carol (MCMC) sampler is used to implement the posterior inference. Compared to other state-of-the-art algorithms, the proposed nonparametric Bayesian algorithm can preserve underlying target clustered properties and effectively suppress these isolated spurious scatterers without any prior information on targets themselves, such as sizes, shapes, and numbers.
- 17 November 2022
- Published by MDPI AG in Applied Sciences
- Vol. 12 (22) , 11684
Open AccessHigh-Resolution Through-the-Wall Radar Imaging with Exploitation of Target Structure
It is quite challenging for through-the-wall radar imaging (TWRI) to achieve high-resolution ghost-free imaging with limited measurements in an indoor multipath scenario. In this paper, a novel high-resolution TWRI algorithm with the exploitation of the target clustered structure in a hierarchical Bayesian framework is proposed. More specifically, an extended spike-and-slab clustered prior is imposed to statistically encourage the cluster formations in both downrange and crossrange domains of the target region, and a generative model of the proposed approach is provided. Then, a Markov Chain Monte Carol (MCMC) sampler is used to implement the posterior inference. Compared to other state-of-the-art algorithms, the proposed nonparametric Bayesian algorithm can preserve underlying target clustered properties and effectively suppress these isolated spurious scatterers without any prior information on targets themselves, such as sizes, shapes, and numbers.