Xiaopeng Zhao

Xiaopeng Zhao

Auto-generated profile, based on publicly available information

Xiaopeng Zhao

Department of Computing, The Hong Kong Polytechnic University, Hong Kong

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
2 October 2023
Verified
Conference Paper
Open access
XZXiaopeng Zhao
Xiaopeng Zhao
ZAZhenlin An
Zhenlin An
QPQingrui Pan
Qingrui Pan
LYLei Yang
Lei Yang
Although Maxwell discovered the physical laws of electromagnetic waves 160 years ago, how to precisely model the propagation of an RF signal in an electrically large and complex environment remains a long-standing problem. The difficulty is in the complex interactions between the RF signal and the obstacles (e.g., reflection, diffraction, etc.). Inspired by the great success of using a neural network to describe the optical field in computer vision, we propose a neural radio-frequency radiance field, NeRF2, which represents a continuous volumetric scene function that makes sense of an RF signal's propagation. Particularly, after training with a few signal measurements, NeRF2 can tell how/what signal is received at any position when it knows the position of a transmitter. As a physical-layer neural network, NeRF2 can take advantage of the learned statistic model plus the physical model of ray tracing to generate a synthetic dataset that meets the training demands of application-layer artificial neural networks (ANNs). Thus, we can boost the performance of ANNs by the proposed turbo-learning, which mixes the true and synthetic datasets to intensify the training. Our experiment results show that turbo-learning can enhance performance with an approximate 50% increase. We also demonstrate the power of NeRF2 in the field of indoor localization and 5G MIMO.

3 June 2024
Verified
Conference Paper
Open access
XZXiaopeng Zhao
Xiaopeng Zhao
GWGuosheng Wang
Guosheng Wang
ZAZhenlin An
Zhenlin An
QPQingrui Pan
Qingrui Pan
LYLei Yang
Lei Yang
Conventional deep learning approaches for indoor localization often suffer from their reliance on high-quality training samples and display limited adaptability across varied scenarios. To address these challenges, we repurpose the Transformer model, celebrated for its profound contextual insights, to explore the underlying principles of indoor localization. Our microbenchmark results compellingly demonstrate the superiority of our approach, showing improvements of 30% to 70% across a diverse set of 50 scenarios compared to other state-of-the-art methods. In conclusion, we propose a specialized Generative Pre-training Transformer (GPT) variant, termed LocGPT, configured with 36 million parameters that are tailored to facilitate transfer learning. By fine-tuning this pre-trained model, we achieve near-par accuracy using merely half the conventional dataset, thereby heralding a pioneering stride in transfer learning within the indoor localization domain.

12 November 2023
Verified
Conference Paper
Open access
SLSicong Liao
Sicong Liao
ZAZhenlin An
Zhenlin An
QPQingrui Pan
Qingrui Pan
XZXiaopeng Zhao
Xiaopeng Zhao
JTJingyu Tong
Jingyu Tong
LYLei Yang
Lei Yang
Cross-Technology Communication (CTC) is an emerging technology that enables physical-layer direct communication from a WiFi sender to other Internet of Things (IoT) receivers via waveform emulation. The previous works use the reverse engineering to find the appropriate WiFi payload that can emulate the waveform similar to the desired IoT packet in the format of the IoT protocol (e.g., ZigBee). Unfortunately, the reverse engineering approach suffers from many limitations, such as being non-reversible and unscalable, misaligning symbols, and over-relying on empiricism. In this work, we present XiTuXi, a one-size-fits-all solution to automatically achieve the CTC by taking advantage of the neural machine translation (NMT), inspired by the task comparability between CTC and homophony-based cross-linguistic communication. We employ a well-known NMT model called Transformer to learn the bit-sequence to bit-sequence translation rationale behind the CTC without human intervention. Particularly, we introduce the forward engineering to address the dilemma of acquiring training datasets. By using XiTuXi, we achieved the CTC with 30 protocol combinations (ie., 802.11b, g, n, ax, ah Å ZigBee, Bluetooth, LoRa, and Sigfox) effortlessly, which ultimately liberates the experts from previous tedious tasks.

22 August 2024 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
XZXiaopeng Zhao
Xiaopeng Zhao
SWShen Wang
Shen Wang
ZAZhenlin An
Zhenlin An
LYLei Yang
Lei Yang
In urban planning and research, 3D city maps are crucial for activities such as cellular network design, urban development, and climate research. Traditionally, creating these models has involved costly techniques like manual 3D mapping, interpretation of satellite or aerial images, or the use of sophisticated depth-sensing equipment. In this work, we propose a novel approach to develop 3D urban maps by examining the influence of urban structures on satellite signals, using GPS records crowdsourced from hundreds of smartphones during everyday user movements. We introduce the concept of satellitic radiance fields (SaRF), a novel neural scene representation technique designed to capture the distribution of GPS signals in urban settings. SaRF employs a sparse voxel octree framework to depict voxel-centric implicit fields, capturing physical properties like the density of each voxel. This model is progressively refined using a differentiable ray-marching process, ultimately leading to the reconstruction of 3D urban maps. Our thorough experimental evaluation, which incorporates approximately 27.4 million GPS records, reveals an average reconstruction accuracy of 83.1% in six varied urban scenes.

4 December 2024
Verified
Conference Paper
Open access
SWShen Wang
Shen Wang
XZXiaopeng Zhao
Xiaopeng Zhao
Donghui DaiDonghui Dai
Donghui Dai
LYLei Yang
Lei Yang
In the era of ubiquitous short-video sharing, an overlooked yet significant privacy risk has arisen: the accidental disclosure of confidential information through reflective surfaces, such as mirrors, glass, or even polished metal. Such reflections can inadvertently disclose sensitive personal details to a broad audience without the awareness of content creators. Our examination of 100 top-rated TikTok short videos reveals that, on average, 37.2% of frames in each video feature identifiable reflective surfaces, posing potential privacy risks. In this work, we introduce a framework designed to scrutinize reflective privacy risks in glass-laden short videos. At the heart of the framework is the development of a reflection-specific neural radiance field, termed RP-NeRF, which enables reflection-aware ray tracing for precise extraction and reconstruction of the reflective scenes from the surfaces they appear on. A detailed field study on the framework indicates that the precision in detecting human presence and recognizing objects from the reconstructed reflective images reaches as high as 90.8% and 89.6%, respectively, even when dealing with a reflective surface that boasts 90% transparency and a mere 4% reflectance rate. These findings highlight the urgent need for greater awareness and advanced solutions to safeguard privacy in our digital age, especially in light of the significant impact of short-video sharing.

3 July 2025 in IEEE Transactions on Mobile Computing
Verified
Journal Article
Closed access
XZXiaopeng Zhao
Xiaopeng Zhao
ZAZhenlin An
Zhenlin An
QPQingrui Pan
Qingrui Pan
LYLei Yang
Lei Yang
Although Maxwell discovered the physical laws of electromagnetic waves about 160 years ago, accurately modeling the propagation of RF signals in large and complex electrical environments remains a persistent challenge. This complexity arises from the interactions between the RF signal and various obstacles, including reflection and diffraction. Inspired by the success of neural networks in mapping the optical field in computer vision, we introduce the neural radio-frequency radiance field, or NeRF$^{2}$2. This represents a continuous volumetric scene function that effectively models RF signal propagation. Remarkably, after only a sparse amount of training with signal measurements, NeRF$^{2}$2 can accurately predict the nature and origin of signals received at any location, assuming the transmitter’s position is known. Additionally, we propose the frequency-aware NeRF$^{2}$2 to enhance channel prediction performance for wideband signals using an RF prism module. Compared to the vanilla NeRF$^{2}$2, the frequency-aware NeRF$^{2}$2 achieves a 4 dB improvement in SNR for FDD OFDM channel estimation and is nearly 3.5 × faster. Functioning as a physical-layer neural network, NeRF$^{2}$2 also supports application-layer artificial neural networks (ANNs) by generating synthetic training datasets. Our empirical results demonstrate that augmented sensing enhances the accuracy of AoA estimation, achieving an approximate 50% improvement.