Zhengyang Su

Zhengyang Su

Auto-generated profile, based on publicly available information

Zhengyang Su

School of Electrical and Electronic Engineering, NTU, Singapore

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 December 2021 in IEEE Transactions on Vehicular Technology
Verified
Journal Article
Closed access
ZSZhengyang Su
Zhengyang Su
KTKah Chan Teh
Kah Chan Teh
SRSirajudeen Gulam Razul
Sirajudeen Gulam Razul
AKAlex Chichung Kot
Alex Chichung Kot
In this paper, we propose a robust non-cooperative spectrum sensing algorithm based on deep learning over Rayleigh fading channel. We conduct noise cancellation on the received sensing data using the stacked convolutional auto-encoder (SCAE) as a pre-processing step. The series of the denoised signal in the time domain is then fed into the proposed Hybrid CNN-SA-GRU (H-CSG) network. The proposed network combines convolutional neural network (CNN), self-attention (SA) modules and gate recurrent unit (GRU). It can extract input features from spatial and temporal domains. The proposed algorithm has been shown to be effective and robust in detecting weak signals at the low signal-to-noise ratio (SNR) level.

16 October 2023 in IEEE Transactions on Vehicular Technology
Verified
Journal Article
Closed access
ZSZhengyang Su
Zhengyang Su
KTKah Chan Teh
Kah Chan Teh
YXYihang Xie
Yihang Xie
SRSirajudeen Gulam Razul
Sirajudeen Gulam Razul
AKAlex C. Kot
Alex C. Kot
Deep learning (DL) has emerged as a promising solution for addressing spectrum scarcity and improving spectrum utilization in cognitive radio networks. Prior to the spectrum sensing phase, additional pre-processing methods could be employed to augment signal quality, while leading to increased computational complexity. In this paper, we introduce a novel joint denoising and spectrum sensing (JDSS) network that leverages an effective loss function to improve the probability of accurate signal detection. The JDSS encompasses a denoising network that regresses noisy received signals and a detection network designed for accurate hypothesis prediction. A comprehensive set of simulations is presented, demonstrating the enhanced performance of our proposed algorithm in a non-cooperative spectrum sensing scenario.

15 December 2022 in IEEE Wireless Communications Letters
Verified
Journal Article
Closed access
ZSZhengyang Su
Zhengyang Su
KTKah Chan Teh
Kah Chan Teh
SRSirajudeen Gulam Razul
Sirajudeen Gulam Razul
AKAlex C. Kot
Alex C. Kot
Deep learning (DL) has been introduced to cognitive radio network to solve the problem of spectrum scarcity and further enhance the spectrum utilization. However, many DL-based spectrum sensing methods are sensitive to the environment, which means the sensing model needs to be re-trained with a large number of labelled samples in a new environment. In this letter, we propose a novel DL-based channel environment-robust spectrum sensing network named ER-SNet, which contains the encoder part extracting channel invariant features and the classifier part for true hypothesis prediction. Extensive simulations have been conducted to show the performance improvement and robustness of the proposed algorithm in sensing weak signals over different channel conditions.

5 June 2024 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
YHYu Hai
Yu Hai
ZSZhengyang Su
Zhengyang Su
JWJunjie Wu
Junjie Wu
KTKah Chan Teh
Kah Chan Teh
ZSZhaoyi Shao
Zhaoyi Shao
JYJianyu Yang
Jianyu Yang
High-resolution Synthetic Aperture Radar (HR-SAR) is extensively used in ground remote sensing applications, including disaster monitoring and resource crop assessment. This is attributed to its exceptional high-resolution imaging capabilities. Microwave photonics technology plays a crucial role in enhancing the performance of SAR systems. It enables the direct emission of ultra-wideband signals across multiple frequency bands. However, this advancement also makes Microwave Photonic SAR (MWP-SAR) susceptible to complex and diverse electromagnetic interference. Particularly, it is vulnerable to radio frequency interference (RFI), and this interference seriously affects the high-resolution imaging results. In order to solve these problems, an MWP-SAR imaging algorithm based on depth spectral sensing and spectral reconstruction that is suitable for complex electromagnetic environments is proposed. Simultaneously, a Sensing-Removing-Recovering (SRR) anti-interference imaging framework is established. To address the high-precision detection of various unknown interferences in MWP-SAR echoes, a deep learning network is constructed. This network is based on spectrum sensing theory and achieves an interference detection probability greater than 98% in various interference scenarios. Subsequently, targeting the continuous spectrum loss in the MWP-SAR signal after interference removal, a signal reconstruction algorithm is proposed. This algorithm employs Toeplitz transformation for signal loss scenarios, facilitating the recovery of signals with extensive spectrum loss post-interference removal. The Factor Group Sparse Regularization (FGSR) algorithm is used to quickly solve the signal recovery problem. Through simulation and measured data processing, the superiority of the proposed algorithm over existing interference suppression imaging algorithms is demonstrated.

20 February 2025 in IEEE Transactions on Vehicular Technology
Verified
Journal Article
Closed access
BRBing Ren
Bing Ren
ZSZhengyang Su
Zhengyang Su
KTKah Chan Teh
Kah Chan Teh
HAHongyang An
Hongyang An
EGErry Gunawan
Erry Gunawan
AKAlex C. Kot
Alex C. Kot
Orthogonal frequency division multiplexing (OFDM) technology, robust against frequency-selective fading channels, is widely applied in the fifth-generation (5G) communication systems, which makes automatic modulation classification (AMC) tasks critical for OFDM communication systems in military and civilian applications. This paper proposes a self-supervised learning framework, named cross-view prediction with automatic augmentation (CPAA). Unlike existing supervised methods requiring substantial labeled data, CPAA leverages massive unlabeled samples and minimal labeled data. Initially, an automatic data augmentation module is incorporated with traditional augmentation methods. Subsequently, the unlabeled pairwise signals are utilized for self-supervised pertaining through the cross-view prediction (CP) module. Finally, supervised fine-tuning using well-trained encoder is conducted on a small amount of labeled datasets. The proposed framework is capable of identifying five different OFDM modulation formats and enhancing performance through automatic data augmentation, which effectively retains data identity and generates diverse views. Simulation results indicate that CPAA can achieve robust classification performance with massive unlabeled samples and extremely limited labeled samples. Moreover, the proposed framework outperforms state-of-the-art methods across a range of signal-to-noise ratio (SNR) levels and demonstrates better generalization capabilities under low SNR conditions. Based on the simulation results, the CPAA framework achieves a classification accuracy of over $\bm {85\%}$ at SNR $\bm {>}$ 8dB with only one labeled sample per SNR for each modulation scheme. Moreover, the automatic data augmentation module enhances the performance of self-supervised learning when limited labeled data are available, yielding an approximate average performance gain of $\bm {48.78\%}$ and demonstrating good transferability on the existing self-supervised learning algorithms.

20 February 2025 in IEEE Transactions on Cognitive Communications and Networking
Verified
Journal Article
Closed access
ZSZhengyang Su
Zhengyang Su
KTKah Chan Teh
Kah Chan Teh
YXYihang Xie
Yihang Xie
SRSirajudeen Gulam Razul
Sirajudeen Gulam Razul
AKAlex C. Kot
Alex C. Kot
Non-cooperative spectrum sensing is a critical aspect of efficient spectrum utilization in wireless communications. However, the reliance on extensive labeled data poses a significant challenge to the application of deep learning methodologies in this domain. To address this issue, we propose a novel contrastive self-supervised learning framework called Dual-Contrast Self-Supervised Spectrum Sensing (DC4S). Inspired by the success of contrastive self-supervised learning in computer vision (CV) and natural language processing (NLP), DC4S extends its application to the field of communications, aiming to reduce the dependency on labeled data while learning robust and discriminative signal representations. DC4S innovatively integrates cross-view prediction and signal-context contrasting to exploit the intrinsic characteristics of communication signals, facilitating robust feature extraction without the need for labeled datasets. The cross-view prediction module leverages the temporal dynamics of communication signals, enabling the model to predict future signal states by understanding the past, thus fostering a deep temporal coherence in the learned representations. Subsequently, the signal-context contrasting module is designed to refine these representations by focusing on the contextual features of the signals, ensuring that the extracted features are discriminatively robust and contextually enriched. Extensive experiments are conducted to evaluate the performance of DC4S with various hyper-parameters on orthogonal frequency division multiplexing (OFDM) signals. The evaluation results demonstrate that fine-tuning a linear classifier in conjunction with the encoder pre-trained in the self-supervised stage achieves comparable performance to supervised learning methods when there exist only 5 labeled samples (1%) per signal-to-noise ratio (SNR) level. Furthermore, DC4S exhibits robust performance in scenarios with limited labeled data and showcases its effectiveness in transfer learning contexts.