Hancong Feng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hancong Feng has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
- Kaili Jiang
University of Electronic Science and Technology of China
9 shared publications - Bin Tang
University of Electronic Science and Technology of China
8 shared publications Bin Tang
School of Information and Communication Engineering University of Electronic Science and Technology of China (UESTC) Chengdu China
5 shared publications- Yuxin Zhao
University of Electronic Science and Technology of China
5 shared publications Kailun Tian
4 shared publications- Kailun Tian
School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China
3 shared publications
Latest Publications
- Open Access
- 1 January 2026
- Published by Institution of Engineering and Technology (IET) in IET Radar, Sonar & Navigation
- Vol. 20 (1)
Sig2text: A Vision‐Language Model for Non‐Cooperative Radar Signal Parsing
Automatic noncooperative analysis of intercepted radar signals is essential for intelligent equipment in both military and civilian domains. Accurate modulation identification and parameter estimation enable effective signal classification, threat assessment and the development of countermeasures. In this paper, we propose a symbolic approach for radar signal recognition and parameter estimation based on a vision‐language model that combines context‐free grammar with time‐frequency representation of radar waveforms. The proposed model, called Sig2text, leverages the power of vision transformers for time‐frequency feature extraction and transformer‐based decoders for symbolic parsing of radar waveforms. By treating radar signal recognition as a parsing problem, Sig2text can recognise the types and estimate the parameters of signals with arbitrarily complex modulations. We evaluate the performance of Sig2text on a synthetic radar signal dataset and demonstrate its effectiveness in various scenarios.
- 1 January 2026
- Published by Institution of Engineering and Technology (IET) in IET Radar, Sonar & Navigation
- Vol. 20 (1)
Open AccessSig2text: A Vision‐Language Model for Non‐Cooperative Radar Signal Parsing
Automatic noncooperative analysis of intercepted radar signals is essential for intelligent equipment in both military and civilian domains. Accurate modulation identification and parameter estimation enable effective signal classification, threat assessment and the development of countermeasures. In this paper, we propose a symbolic approach for radar signal recognition and parameter estimation based on a vision‐language model that combines context‐free grammar with time‐frequency representation of radar waveforms. The proposed model, called Sig2text, leverages the power of vision transformers for time‐frequency feature extraction and transformer‐based decoders for symbolic parsing of radar waveforms. By treating radar signal recognition as a parsing problem, Sig2text can recognise the types and estimate the parameters of signals with arbitrarily complex modulations. We evaluate the performance of Sig2text on a synthetic radar signal dataset and demonstrate its effectiveness in various scenarios.
- 18 June 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 293
Deep multi-intentional inverse reinforcement learning for cognitive multi-function radar inverse cognition
- 18 June 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 293
Deep multi-intentional inverse reinforcement learning for cognitive multi-function radar inverse cognition
- 18 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 61 (4) , 8143-8158
Joint Pulse Association, Sorting, and Localization of Pulse Radar Signals With Cluster Reconnaissance
A significant challenge in cluster collaborative reconnaissance is achieving pulse association between disparate reconnaissance units. However, traditional time-of-arrival (TOA) histogram methods struggle to attain high accuracy in radar pulse association when reconnaissance units are distant, the pulse repetition frequency (PRF) of the radar signal is high, or the emitter's position is suboptimal. Furthermore, these methods are sensitive to the division of the TOA histogram. To address these issues, we propose a pulse association method for cluster reconnaissance that treats the association, sorting, and localization of radar signals as a unified problem, aiming to solve them collectively. Additionally, we introduce the concept of a fuzzy generation set to resolve ambiguities in localization. Simulation results demonstrate that our proposed method significantly enhances the accuracy of pulse association while also achieving high sorting accuracy and localization precision. Under conditions with a TOA error of 50 ns, a site error of 10 m and a pulse drop rate of 0.2, our proposed method achieves an estimation accuracy exceeding 90% for the emitters count, with pulse association and sorting accuracies greater than 95%, and a localization RMSE of less than 1%.
- 18 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 61 (4) , 8143-8158
Joint Pulse Association, Sorting, and Localization of Pulse Radar Signals With Cluster Reconnaissance
A significant challenge in cluster collaborative reconnaissance is achieving pulse association between disparate reconnaissance units. However, traditional time-of-arrival (TOA) histogram methods struggle to attain high accuracy in radar pulse association when reconnaissance units are distant, the pulse repetition frequency (PRF) of the radar signal is high, or the emitter's position is suboptimal. Furthermore, these methods are sensitive to the division of the TOA histogram. To address these issues, we propose a pulse association method for cluster reconnaissance that treats the association, sorting, and localization of radar signals as a unified problem, aiming to solve them collectively. Additionally, we introduce the concept of a fuzzy generation set to resolve ambiguities in localization. Simulation results demonstrate that our proposed method significantly enhances the accuracy of pulse association while also achieving high sorting accuracy and localization precision. Under conditions with a TOA error of 50 ns, a site error of 10 m and a pulse drop rate of 0.2, our proposed method achieves an estimation accuracy exceeding 90% for the emitters count, with pulse association and sorting accuracies greater than 95%, and a localization RMSE of less than 1%.
- 18 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 61 (2) , 1901-1913
Deinterleaving of Pulse Streams With Conditional Autoregressive Kernel Mixture Network
Deinterleaving emitters with complex patterns presents a significant challenge for electronic support measure (ESM) systems. In this paper, we address this issue by formulating deinterleaving as the optimization of an autoregressive likelihood function. We then propose the Conditional Autoregressive Kernel Mixture Network (CARKMN), a conditional generative model that estimates the conditional density of pulse parameters based on previous noisy observations and source labels to estimate the solution of this optimization problem. The model, trained with a modified loss function for denoising, can extract pulses belonging to the class of a given source label. Therefore, the denoising-based deinterleaving is achieved with a single model. We evaluate our model with the proposed algorithms on a challenging synthetic dataset under various non-ideal conditions and compare it against existing approaches for both conventional and open-set deinterleaving. The results indicate that our method significantly outperforms the comparative techniques, especially in open-set deinterleaving.
- 18 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 61 (2) , 1901-1913
Deinterleaving of Pulse Streams With Conditional Autoregressive Kernel Mixture Network
Deinterleaving emitters with complex patterns presents a significant challenge for electronic support measure (ESM) systems. In this paper, we address this issue by formulating deinterleaving as the optimization of an autoregressive likelihood function. We then propose the Conditional Autoregressive Kernel Mixture Network (CARKMN), a conditional generative model that estimates the conditional density of pulse parameters based on previous noisy observations and source labels to estimate the solution of this optimization problem. The model, trained with a modified loss function for denoising, can extract pulses belonging to the class of a given source label. Therefore, the denoising-based deinterleaving is achieved with a single model. We evaluate our model with the proposed algorithms on a challenging synthetic dataset under various non-ideal conditions and compare it against existing approaches for both conventional and open-set deinterleaving. The results indicate that our method significantly outperforms the comparative techniques, especially in open-set deinterleaving.
- Open Access
- 1 May 2024
- Published by KEAI Publishing in Defence Technology
- Vol. 35, 275-288
Probabilistic modeling of multifunction radars with autoregressive kernel mixture network
- 1 May 2024
- Published by KEAI Publishing in Defence Technology
- Vol. 35, 275-288
Open AccessProbabilistic modeling of multifunction radars with autoregressive kernel mixture network
- 22 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 60 (4) , 5060-5072
Syntactic Modeling and Neural-Based Parsing for Multifunction Radar Signal Interpretation
The analysis of intercepted multifunction radar (MFR) signals and the assessment of potential threats are critical in the field of radar countermeasures. The complexity of MFR control mechanisms poses serious challenges in modeling and recognizing their internal states. MFRs, being discrete event systems (DESs), can be well described by formal languages. This paper presents a generalized version of the previous syntactic modeling of MFRs, utilizing the proposed latent variable controlled stochastic context-free grammar (LVCSCFG). The enhanced model incorporates a mapping between the external environment and the production probabilities, offering a more effective representation of the resource management process. Additionally, we introduce a neural-based approach for maximum likelihood estimation of the MFR's internal states, which are represented by a grammar tree. LVCSCFG exhibits enhanced representation of MFR dynamic behaviors, and simulation results indicate that the neural network outperforms traditional statistical parsing algorithms in terms of speed and precision under various non-ideal channel conditions.
- 22 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 60 (4) , 5060-5072
Syntactic Modeling and Neural-Based Parsing for Multifunction Radar Signal Interpretation
The analysis of intercepted multifunction radar (MFR) signals and the assessment of potential threats are critical in the field of radar countermeasures. The complexity of MFR control mechanisms poses serious challenges in modeling and recognizing their internal states. MFRs, being discrete event systems (DESs), can be well described by formal languages. This paper presents a generalized version of the previous syntactic modeling of MFRs, utilizing the proposed latent variable controlled stochastic context-free grammar (LVCSCFG). The enhanced model incorporates a mapping between the external environment and the production probabilities, offering a more effective representation of the resource management process. Additionally, we introduce a neural-based approach for maximum likelihood estimation of the MFR's internal states, which are represented by a grammar tree. LVCSCFG exhibits enhanced representation of MFR dynamic behaviors, and simulation results indicate that the neural network outperforms traditional statistical parsing algorithms in terms of speed and precision under various non-ideal channel conditions.
- Open Access
- 12 April 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (8) , 1370
Two-Dimensional Target Localization Approach via a Closed-Form Solution Using Range Difference Measurements Based on Pentagram Array
This paper presents a simple and fast closed-form solution approach for two-dimensional (2D) target localization using range difference (RD) measurements. The formulation of the localization problem is derived using a pentagram array. The target position is determined using passive radar measurements (RDs) between the target and the () receivers’ locations. The method facilitates the problem of target position and can be used as a counter-parallel method for spherical interpolation (SI) and spherical intersection (SX) methods in time difference of arrival (TDOA) and radar systems. The performance of the method is examined in 2D target localization using numerical analysis under the distribution of receivers in the pentagram array. The simulations are conducted using four different far-distance targets and comparatively large-area distributed receivers. The RD measurements were distorted by two different values of Gaussian errors based on ionosphedriec time delays of 20 and 50 nsec owing to the different receivers’ positions. The findings highly verified the validity of the method for addressing the problem of target localization. Additionally, a theoretical accuracy study of the method is given, which solely relies on the RD measurements.- 12 April 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (8) , 1370
Open AccessTwo-Dimensional Target Localization Approach via a Closed-Form Solution Using Range Difference Measurements Based on Pentagram Array
This paper presents a simple and fast closed-form solution approach for two-dimensional (2D) target localization using range difference (RD) measurements. The formulation of the localization problem is derived using a pentagram array. The target position is determined using passive radar measurements (RDs) between the target and the () receivers’ locations. The method facilitates the problem of target position and can be used as a counter-parallel method for spherical interpolation (SI) and spherical intersection (SX) methods in time difference of arrival (TDOA) and radar systems. The performance of the method is examined in 2D target localization using numerical analysis under the distribution of receivers in the pentagram array. The simulations are conducted using four different far-distance targets and comparatively large-area distributed receivers. The RD measurements were distorted by two different values of Gaussian errors based on ionosphedriec time delays of 20 and 50 nsec owing to the different receivers’ positions. The findings highly verified the validity of the method for addressing the problem of target localization. Additionally, a theoretical accuracy study of the method is given, which solely relies on the RD measurements. - Open Access
- 26 February 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (5) , 811
A Fast Power Spectrum Sensing Solution for Generalized Coprime Sampling
With the growing scarcity of spectrum resources, wideband spectrum sensing is necessary to process a large volume of data at a high sampling rate. For some applications, only second-order statistics are required for spectrum estimation. In this case, a fast power spectrum sensing solution is proposed based on the generalized coprime sampling. The solution involves the inherent structure of the sensing vector to reconstruct the autocorrelation sequence of inputs from sub-Nyquist samples, which requires only parallel Fourier transform and simple multiplication operations. Thus, it takes less time than the state-of-the-art methods while maintaining the same performance, and it achieves higher performance than the existing methods within the same execution time without the need to pre-estimate the number of inputs. Furthermore, the influence of the model mismatch has only a minor impact on the estimation performance, allowing for more efficient use of the spectrum resource in a distributed swarm scenario. Simulation results demonstrate the low complexity in sampling and computation, thus making it a more practical solution for real-time and distributed wideband spectrum sensing applications.
- 26 February 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (5) , 811
Open AccessA Fast Power Spectrum Sensing Solution for Generalized Coprime Sampling
With the growing scarcity of spectrum resources, wideband spectrum sensing is necessary to process a large volume of data at a high sampling rate. For some applications, only second-order statistics are required for spectrum estimation. In this case, a fast power spectrum sensing solution is proposed based on the generalized coprime sampling. The solution involves the inherent structure of the sensing vector to reconstruct the autocorrelation sequence of inputs from sub-Nyquist samples, which requires only parallel Fourier transform and simple multiplication operations. Thus, it takes less time than the state-of-the-art methods while maintaining the same performance, and it achieves higher performance than the existing methods within the same execution time without the need to pre-estimate the number of inputs. Furthermore, the influence of the model mismatch has only a minor impact on the estimation performance, allowing for more efficient use of the spectrum resource in a distributed swarm scenario. Simulation results demonstrate the low complexity in sampling and computation, thus making it a more practical solution for real-time and distributed wideband spectrum sensing applications.
- Open Access
- 31 January 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (3) , 535
Pentagram Arrays: A New Paradigm for DOA Estimation of Wideband Sources Based on Triangular Geometry
Antenna arrays are used for signal processing in sonar and radar direction of arrival (DOA) estimation. The well-known array geometries used in DOA estimation are uniform linear array (ULA), uniform circular array (UCA), and rectangular grid array (RGA). In these geometries, the neighboring elements are separated by a fixed distance λ/2 (λ is the wavelength), which does not perform well for d greater than λ/2. Uniform rectangular arrays introduce grating lobes, which cause poor DOA estimation performance, especially for wideband sources. Random sampling arrays are sometimes practically not realizable. Periodic geometries require numerous sensors. Based on the minimization of the number of sensors, this paper developed a novel pentagram array to address the problem of DOA estimation of wideband sources. The array has a fixed number of elements with variable element spacing and is abbreviated as (FNEVES), which offers a new idea for array design. In this study, the geometric structure is designed and mathematically analyzed. Also, a DOA signal model is designed based on a spherical radar coordinate system to derive its steering manifold matrix. The DOA estimation performance comparison with ULA and UCA geometries under the multiple signal classification (MUSIC) algorithm using different wideband scenarios is presented. For further investigation, more simulations are realized using the minimum variance distortionless (MVDR) technique (CAPON) and the subtracting signal subspace (SSS) algorithm. Simulation results demonstrate the effectiveness of the proposed geometry compared to its counterparts. In addition, the SSS, through the simulations, provided better results than the MUSIC and CAPON methods.
- 31 January 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (3) , 535
Open AccessPentagram Arrays: A New Paradigm for DOA Estimation of Wideband Sources Based on Triangular Geometry
Antenna arrays are used for signal processing in sonar and radar direction of arrival (DOA) estimation. The well-known array geometries used in DOA estimation are uniform linear array (ULA), uniform circular array (UCA), and rectangular grid array (RGA). In these geometries, the neighboring elements are separated by a fixed distance λ/2 (λ is the wavelength), which does not perform well for d greater than λ/2. Uniform rectangular arrays introduce grating lobes, which cause poor DOA estimation performance, especially for wideband sources. Random sampling arrays are sometimes practically not realizable. Periodic geometries require numerous sensors. Based on the minimization of the number of sensors, this paper developed a novel pentagram array to address the problem of DOA estimation of wideband sources. The array has a fixed number of elements with variable element spacing and is abbreviated as (FNEVES), which offers a new idea for array design. In this study, the geometric structure is designed and mathematically analyzed. Also, a DOA signal model is designed based on a spherical radar coordinate system to derive its steering manifold matrix. The DOA estimation performance comparison with ULA and UCA geometries under the multiple signal classification (MUSIC) algorithm using different wideband scenarios is presented. For further investigation, more simulations are realized using the minimum variance distortionless (MVDR) technique (CAPON) and the subtracting signal subspace (SSS) algorithm. Simulation results demonstrate the effectiveness of the proposed geometry compared to its counterparts. In addition, the SSS, through the simulations, provided better results than the MUSIC and CAPON methods.
- 13 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 59 (6) , 9621-9634
Self-Supervised Contrastive Learning for Extracting Radar Word in the Hierarchical Model of Multifunction Radar
The analysis of intercepted multifunction radar (MFR) signals has attracted considerable attention in the field of cognitive electronic reconnaissance. The agility of pulse parameters makes it difficult to recognize their behavior states. Currently, most existing approaches for recognizing MFR behaviors heavily depend on prior information, which can hardly be obtained in a non cooperative way. This study develops a contrastive self supervised method to adaptively segment and cluster MFR pulse sequences into radar words with little prior information. First, a convolutional neural network (CNN) is trained in a self-supervised manner to differentiate between adjacent and nonadjacent pulse group embeddings to learn robust feature representations from sequences of pulse descriptor words (PDWs). Next, based on the learned embeddings, test statistics are calculated for each PDW sequence to detect change points, which are in turn used to segment the sequence into multiple subsegments, each containing only one radar word. Finally, feature vectors are produced for clustering the subsegments into groups of radar words. Simulation results show that without using any labeled data, the proposed method can effectively extract the radar words of a hypothetical MFR under corrupted and overlapped pulse parameters and performs only slightly worse than the state-of-the-art fully supervised method.
- 13 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 59 (6) , 9621-9634
Self-Supervised Contrastive Learning for Extracting Radar Word in the Hierarchical Model of Multifunction Radar
The analysis of intercepted multifunction radar (MFR) signals has attracted considerable attention in the field of cognitive electronic reconnaissance. The agility of pulse parameters makes it difficult to recognize their behavior states. Currently, most existing approaches for recognizing MFR behaviors heavily depend on prior information, which can hardly be obtained in a non cooperative way. This study develops a contrastive self supervised method to adaptively segment and cluster MFR pulse sequences into radar words with little prior information. First, a convolutional neural network (CNN) is trained in a self-supervised manner to differentiate between adjacent and nonadjacent pulse group embeddings to learn robust feature representations from sequences of pulse descriptor words (PDWs). Next, based on the learned embeddings, test statistics are calculated for each PDW sequence to detect change points, which are in turn used to segment the sequence into multiple subsegments, each containing only one radar word. Finally, feature vectors are produced for clustering the subsegments into groups of radar words. Simulation results show that without using any labeled data, the proposed method can effectively extract the radar words of a hypothetical MFR under corrupted and overlapped pulse parameters and performs only slightly worse than the state-of-the-art fully supervised method.