Zhaoyang Qiu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zhaoyang Qiu has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
Tianyun Li
School of Information Engineering, Information Engineering University, Zhengzhou, China
4 shared publications- Xiong Zha
School of Information Engineering, Information Engineering University, Zhengzhou, China
3 shared publications - Chen Yi
School of Communications and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing, China
2 shared publications - Tianyun Li
School of Naval Architecture and Ocean Engineering, Huazhong University of Science and Technology, Wuhan 430074, P. R. China
2 shared publications - Yong Li
College of Computer Science, Chongqing University, Chongqing, China
2 shared publications Bingxi Song
Hebei Key Laboratory of Electromagnetic Spectrum Cognition and Control, Shijiazhuang, China
1 shared publication
Latest Publications
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 12, 5704-5717
Single-Channel Blind Source Separation of Co-Channel Communication Signals: A Hybrid Knowledge-Data Driven Approach
Single-channel blind source separation (SCBSS) aims to recover multiple source signals from a single-channel mixture of signals. Traditional maximum likelihood methods incur prohibitively computational complexity, while existing deep learning approaches often overlook domain-specific knowledge of communication, which limits their separation accuracy. To address this gap, we propose a complex-valued convolutional encoder-separator-decoder network (C2ESDNet) for efficient high-precision SCBSS of co-channel communication signals. Unlike real-valued approximations, our complex-valued network inherently models in-phase and quadrature components without information loss. The encoder-separator-decoder architecture conforms to the weighted superposition principle of SCBSS. Furthermore, stacked dilated convolution blocks are employed to capture multi-scale waveform features. More importantly, we analyze and prove that the conventional scale-invariant source-to-noise (SISNR) ratio loss fails to preserve amplitude-phase characteristics essential for demodulation. Therefore, we propose a demodulation-aware loss, where the mean squared error focuses on point-wise accuracy and SISNR enhances global phase fidelity. Additionally, we further extend C2ESDNet to symbol-aware C2ESDNet-S, which integrates waveform separation with downstream demodulation. It features an extended decoder that mirrors the synchronization-and-demodulation pipeline of communication receivers in a data-driven manner. Numerical results show that our methods achieve a lower bit error rate compared to state-of-the-art (SOTA) baselines. The visualization result shows superior waveform fidelity in the separated signals of C2ESDNet compared to the SOTA baseline. Ablation studies confirm the critical contribution of the demodulation-aware loss to the performance gains.
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 12, 5704-5717
Single-Channel Blind Source Separation of Co-Channel Communication Signals: A Hybrid Knowledge-Data Driven Approach
Single-channel blind source separation (SCBSS) aims to recover multiple source signals from a single-channel mixture of signals. Traditional maximum likelihood methods incur prohibitively computational complexity, while existing deep learning approaches often overlook domain-specific knowledge of communication, which limits their separation accuracy. To address this gap, we propose a complex-valued convolutional encoder-separator-decoder network (C2ESDNet) for efficient high-precision SCBSS of co-channel communication signals. Unlike real-valued approximations, our complex-valued network inherently models in-phase and quadrature components without information loss. The encoder-separator-decoder architecture conforms to the weighted superposition principle of SCBSS. Furthermore, stacked dilated convolution blocks are employed to capture multi-scale waveform features. More importantly, we analyze and prove that the conventional scale-invariant source-to-noise (SISNR) ratio loss fails to preserve amplitude-phase characteristics essential for demodulation. Therefore, we propose a demodulation-aware loss, where the mean squared error focuses on point-wise accuracy and SISNR enhances global phase fidelity. Additionally, we further extend C2ESDNet to symbol-aware C2ESDNet-S, which integrates waveform separation with downstream demodulation. It features an extended decoder that mirrors the synchronization-and-demodulation pipeline of communication receivers in a data-driven manner. Numerical results show that our methods achieve a lower bit error rate compared to state-of-the-art (SOTA) baselines. The visualization result shows superior waveform fidelity in the separated signals of C2ESDNet compared to the SOTA baseline. Ablation studies confirm the critical contribution of the demodulation-aware loss to the performance gains.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 73 (12) , 12993-13004
Identification of the Whole Interleaving Pattern From Randomly Interleaved Linear Block Codes With Known Minimum Distances
In this paper, an effective approach is proposed to recognize the order among codewords from the randomly binary and non-binary interleaved linear block coded streams if the code minimum distance is known, where the interleaver length Nn + d is not the multiple of the linear block code length n. Then, an efficient approach is developed without knowing linear block code parameters such as code rate and the code generator matrix (except the code length) to recognize the whole random interleaving pattern if the greatest common divisor (GCD) of n and d is GCD(n, d) = 1 for both binary and non-binary interleaved linear block coded symbols. Further, for the case GCD(n, d) = q ̸= 1, the whole random interleaving pattern can be recognized with the probability of no more than (q!)−(Nn+d)/q for the following cases: (1) the non-binary interleaved coded symbols without knowing the non-binary code generator matrix; (2) the binary interleaved coded symbols.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 73 (12) , 12993-13004
Identification of the Whole Interleaving Pattern From Randomly Interleaved Linear Block Codes With Known Minimum Distances
In this paper, an effective approach is proposed to recognize the order among codewords from the randomly binary and non-binary interleaved linear block coded streams if the code minimum distance is known, where the interleaver length Nn + d is not the multiple of the linear block code length n. Then, an efficient approach is developed without knowing linear block code parameters such as code rate and the code generator matrix (except the code length) to recognize the whole random interleaving pattern if the greatest common divisor (GCD) of n and d is GCD(n, d) = 1 for both binary and non-binary interleaved linear block coded symbols. Further, for the case GCD(n, d) = q ̸= 1, the whole random interleaving pattern can be recognized with the probability of no more than (q!)−(Nn+d)/q for the following cases: (1) the non-binary interleaved coded symbols without knowing the non-binary code generator matrix; (2) the binary interleaved coded symbols.
- 14 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 72 (10) , 6001-6013
Error-Correcting Codes With Large Field Size Under Non-Binary Segmented Burst Deletion/Insertion Channels and Unknown Codeword Boundaries
In this paper, we construct non-binary codes of length N which correct errors under a non-binary segment- N maxburst- D deletion and maxburst- S insertion (NB-SBDI( N, D, S )) channel without knowing the codeword boundaries. In this NB-SBDI( N, D, S ) channel, at most a single non-binary burst (a block of consecutive bits/symbols) of deletions or insertions of length up to D or S , respectively, exists in a block of N consecutive non-binary symbols. One code named as BM-DB-MDS consists of a maximum distance separable (MDS) code, a block of periodic de Bruijn (DB) symbols, and a block of proposed periodic binary marker (BM) patterns with a period of S + D + 2. The other code called BM-MDS code consists of a BM code and an MDS code. We show that the rates of BM-DB-MDS and BM-MDS codes achieve λ/λ+1 (1 - 1/2 t ), λ ∈ N + and 1 - 1/ t , respectively, when N → +∞, where λ represent the MDS code shortening factor, and 1/ t is the rough proportion of the maximum length of burst deletions or insertions allowed in a code.
- 14 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 72 (10) , 6001-6013
Error-Correcting Codes With Large Field Size Under Non-Binary Segmented Burst Deletion/Insertion Channels and Unknown Codeword Boundaries
In this paper, we construct non-binary codes of length N which correct errors under a non-binary segment- N maxburst- D deletion and maxburst- S insertion (NB-SBDI( N, D, S )) channel without knowing the codeword boundaries. In this NB-SBDI( N, D, S ) channel, at most a single non-binary burst (a block of consecutive bits/symbols) of deletions or insertions of length up to D or S , respectively, exists in a block of N consecutive non-binary symbols. One code named as BM-DB-MDS consists of a maximum distance separable (MDS) code, a block of periodic de Bruijn (DB) symbols, and a block of proposed periodic binary marker (BM) patterns with a period of S + D + 2. The other code called BM-MDS code consists of a BM code and an MDS code. We show that the rates of BM-DB-MDS and BM-MDS codes achieve λ/λ+1 (1 - 1/2 t ), λ ∈ N + and 1 - 1/ t , respectively, when N → +∞, where λ represent the MDS code shortening factor, and 1/ t is the rough proportion of the maximum length of burst deletions or insertions allowed in a code.
- 23 February 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Letters
- Vol. 27 (4) , 1170-1174
Unsupervised Radio Frequency Fingerprint Identification Based on Curriculum Learning
Radio frequency fingerprint (RFF) identification aims to identify emitters by extracting the inherent physical-layer features. Despite the significant accuracy achieved in a supervised learning manner, few works have focused on unsupervised RFF identification. To this end, we propose an unsupervised training paradigm based on the curriculum learning framework. Specifically, the unlabeled samples are first sorted by the signal-to-noise (SNR) of signals. Then, the network is pretrained with contrastive learning on the high SNR subset. To further improve identification performance, the model is gradually optimized from the easy (high SNR) to the hard (low SNR) subsets with pseudo label learning. Experimental results indicate that the proposed scheme achieves comparable performance against schemes based on supervised learning.
- 23 February 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Letters
- Vol. 27 (4) , 1170-1174
Unsupervised Radio Frequency Fingerprint Identification Based on Curriculum Learning
Radio frequency fingerprint (RFF) identification aims to identify emitters by extracting the inherent physical-layer features. Despite the significant accuracy achieved in a supervised learning manner, few works have focused on unsupervised RFF identification. To this end, we propose an unsupervised training paradigm based on the curriculum learning framework. Specifically, the unlabeled samples are first sorted by the signal-to-noise (SNR) of signals. Then, the network is pretrained with contrastive learning on the high SNR subset. To further improve identification performance, the model is gradually optimized from the easy (high SNR) to the hard (low SNR) subsets with pseudo label learning. Experimental results indicate that the proposed scheme achieves comparable performance against schemes based on supervised learning.
- 1 February 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 30 (10709908) , 70-74
Cross-Receiver Radio Frequency Fingerprint Identification Based on Contrastive Learning and Subdomain Adaptation
Radio frequency fingerprint (RFF) identification is emerging as an attractive paradigm for physical layer security. Despite the exceptional accuracy achieved by deep learning (DL) based schemes, few works consider the cross-receiver scenario. The performance deteriorates significantly when the model is deployed on new receivers directly. To this end, a cross-receiver RFF learning scheme is proposed. First, an unsupervised pre-training method based on contrastive learning is utilized to extract receiver-agnostic features. Then, the model is optimized by subdomain adaptation to further improve identification performance. The proposed scheme does not require multiple labeled datasets from different receivers. And experimental results indicate that the proposed scheme effectively alleviates performance degradation in the cross-receiver scenario.
- 1 February 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 30 (10709908) , 70-74
Cross-Receiver Radio Frequency Fingerprint Identification Based on Contrastive Learning and Subdomain Adaptation
Radio frequency fingerprint (RFF) identification is emerging as an attractive paradigm for physical layer security. Despite the exceptional accuracy achieved by deep learning (DL) based schemes, few works consider the cross-receiver scenario. The performance deteriorates significantly when the model is deployed on new receivers directly. To this end, a cross-receiver RFF learning scheme is proposed. First, an unsupervised pre-training method based on contrastive learning is utilized to extract receiver-agnostic features. Then, the model is optimized by subdomain adaptation to further improve identification performance. The proposed scheme does not require multiple labeled datasets from different receivers. And experimental results indicate that the proposed scheme effectively alleviates performance degradation in the cross-receiver scenario.
- 13 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Letters
- Vol. 26 (3) , 592-596
Specific Emitter Identification Based on Complex Fourier Neural Network
Specific emitter identification (SEI) is a well-established approach to providing precise target information for civilian and military applications. For most deep learning (DL) based SEI schemes, neural operators directly learn mappings from the raw baseband waveform or its transformed representation. Different from existing schemes, we propose a novel complex Fourier neural operator (CFNO) in this letter, which introduces a time and frequency domain attention mechanism. With the CFNO block, features are fully learned from different domain perspectives. We evaluate the proposed method based on the joint distortion model of the transmitter and compare it with several state-of-the-art SEI algorithms. Simulation results demonstrate its excellent performance, making the CFNO block a good candidate for extracting fingerprints.
- 13 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Letters
- Vol. 26 (3) , 592-596
Specific Emitter Identification Based on Complex Fourier Neural Network
Specific emitter identification (SEI) is a well-established approach to providing precise target information for civilian and military applications. For most deep learning (DL) based SEI schemes, neural operators directly learn mappings from the raw baseband waveform or its transformed representation. Different from existing schemes, we propose a novel complex Fourier neural operator (CFNO) in this letter, which introduces a time and frequency domain attention mechanism. With the CFNO block, features are fully learned from different domain perspectives. We evaluate the proposed method based on the joint distortion model of the transmitter and compare it with several state-of-the-art SEI algorithms. Simulation results demonstrate its excellent performance, making the CFNO block a good candidate for extracting fingerprints.
- Open Access
- 1 August 2020
- Published by Institution of Engineering and Technology (IET) in IET Communications
- Vol. 14 (14) , 2360-2368
Digital blind despreading method for intermediate frequency short‐code DSSS signal
Direct sequence spread spectrum (DSSS) signals are widely used in various military and civilian communication systems owing to its useful properties, such as lower working signal‐to‐noise ratio, strong anti‐jamming capability, and robustness to multi‐path fading. In past few years, the estimation for spreading sequence in non‐cooperative context has attracted lots of attention. Several methods are developed to realise a recovery of the sequence. Eigenvalue decomposition (EVD) is a classical method to obtain an effective estimation by extracting the main eigenvectors. The only obstacle to apply EVD method into practical systems is high complexity of computation when spreading sequence is long. Alternative algorithms such as neural network and clustering method have lower complexity, but an evident loss in performance. In this work, the authors propose to generalise the EVD algorithm to intermediate frequency (IF) short‐code DSSS with an unknown carrier offset. The blind despreading process is designed to complete correlation demodulation. As a result, the computational afford is largely decreased due to the operation on IF real signal. To evaluate the performance, they derive the Crammer‐Rao bound for spreading code estimation and compare the performance of several methods. Simulation demonstrates the superiority of the proposed method in estimation accuracy and runtime.
- 1 August 2020
- Published by Institution of Engineering and Technology (IET) in IET Communications
- Vol. 14 (14) , 2360-2368
Open AccessDigital blind despreading method for intermediate frequency short‐code DSSS signal
Direct sequence spread spectrum (DSSS) signals are widely used in various military and civilian communication systems owing to its useful properties, such as lower working signal‐to‐noise ratio, strong anti‐jamming capability, and robustness to multi‐path fading. In past few years, the estimation for spreading sequence in non‐cooperative context has attracted lots of attention. Several methods are developed to realise a recovery of the sequence. Eigenvalue decomposition (EVD) is a classical method to obtain an effective estimation by extracting the main eigenvectors. The only obstacle to apply EVD method into practical systems is high complexity of computation when spreading sequence is long. Alternative algorithms such as neural network and clustering method have lower complexity, but an evident loss in performance. In this work, the authors propose to generalise the EVD algorithm to intermediate frequency (IF) short‐code DSSS with an unknown carrier offset. The blind despreading process is designed to complete correlation demodulation. As a result, the computational afford is largely decreased due to the operation on IF real signal. To evaluate the performance, they derive the Crammer‐Rao bound for spreading code estimation and compare the performance of several methods. Simulation demonstrates the superiority of the proposed method in estimation accuracy and runtime.
- 1 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 645-649
Direct Sequence Spread Spectrum Signal Detection in DVB-CID Signal
Carrier Identification(CID) technique is an innovative technology that has solved the problem of radio frequency interference in satellite communication system. In this paper, the detection of CID signal in digital video broadcasting of a carrier identification system(DVB-CID) is deeply studied. We propose a quadratic spectrum detection algorithm based on constellation elimination processing which shows more promising when compared to more traditional reconstruction cancellation processing. Furthermore, we discuss the impact of frequency offset and low signal to noise ratio(SNR) on detection performance though the detection threshold. The CID signal detection can be completed when signal-to-noise ratio is -10dB. Moreover, the method based on constellation elimination processing can eliminate the interference of the main carrier signal to the CID signal more completely, which reduces the complexity of the algorithm and is easy to implement in engineering. The detection threshold of the CID signal can reach 0.3 when SIR is 3 dB.
- 1 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 645-649
Direct Sequence Spread Spectrum Signal Detection in DVB-CID Signal
Carrier Identification(CID) technique is an innovative technology that has solved the problem of radio frequency interference in satellite communication system. In this paper, the detection of CID signal in digital video broadcasting of a carrier identification system(DVB-CID) is deeply studied. We propose a quadratic spectrum detection algorithm based on constellation elimination processing which shows more promising when compared to more traditional reconstruction cancellation processing. Furthermore, we discuss the impact of frequency offset and low signal to noise ratio(SNR) on detection performance though the detection threshold. The CID signal detection can be completed when signal-to-noise ratio is -10dB. Moreover, the method based on constellation elimination processing can eliminate the interference of the main carrier signal to the CID signal more completely, which reduces the complexity of the algorithm and is easy to implement in engineering. The detection threshold of the CID signal can reach 0.3 when SIR is 3 dB.
- Open Access
- 2 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 125700-125710
A Blind Despreading and Demodulation Method for QPSK-DSSS Signal With Unknown Carrier Offset Based on Matrix Subspace Analysis
Direct sequence spread spectrum (DSSS) signals are widely used in various military and civilian communication systems. It adopts the pseudorandom (PN) sequence to modulate baseband signal before transmission, leading to some unique and useful properties such as lower working signal to noise ratio (SNR), strong anti-jamming capability, and robustness to the effect of multi-path fading. In past few years, blind despreading for DSSS signals has been a hot topic in cognitive radio field. Previous works are mostly concentrated on BPSK-DSSS signals, and have developed theories and methods to estimate the PN sequence effectively. But unfortunately, most of them are powerless to process QPSK-DSSS signals. This work proposes a method based on matrix subspace analysis to recover the PN sequences and transmitted symbols for QPSK-DSSS signals with unknown carrier offset under non-cooperative reception occasion. To recover spreading sequences from subspace of self-covariance matrix, we derive the theoretical expressions of eigenvector by studying the complex hermit matrix decomposition and then investigate the influence of carrier residual. Finally, a structure is designed to complete blind despreading and demodulation from the received waveform to bit stream. Simulation verifies pretty good bit-error-rate (BER) performance of proposed algorithm.
- 2 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 125700-125710
Open AccessA Blind Despreading and Demodulation Method for QPSK-DSSS Signal With Unknown Carrier Offset Based on Matrix Subspace Analysis
Direct sequence spread spectrum (DSSS) signals are widely used in various military and civilian communication systems. It adopts the pseudorandom (PN) sequence to modulate baseband signal before transmission, leading to some unique and useful properties such as lower working signal to noise ratio (SNR), strong anti-jamming capability, and robustness to the effect of multi-path fading. In past few years, blind despreading for DSSS signals has been a hot topic in cognitive radio field. Previous works are mostly concentrated on BPSK-DSSS signals, and have developed theories and methods to estimate the PN sequence effectively. But unfortunately, most of them are powerless to process QPSK-DSSS signals. This work proposes a method based on matrix subspace analysis to recover the PN sequences and transmitted symbols for QPSK-DSSS signals with unknown carrier offset under non-cooperative reception occasion. To recover spreading sequences from subspace of self-covariance matrix, we derive the theoretical expressions of eigenvector by studying the complex hermit matrix decomposition and then investigate the influence of carrier residual. Finally, a structure is designed to complete blind despreading and demodulation from the received waveform to bit stream. Simulation verifies pretty good bit-error-rate (BER) performance of proposed algorithm.