Sanxing Cao

Prof. Dr. Sanxing Cao


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
1 January 2023 in Computer Modeling in Engineering & Sciences
Verified
Journal Article
Open access
PDF file
ZXZe Xu
Ze Xu
SCSanxing Cao
Sanxing Cao
Multi-Source data plays an important role in the evolution of media convergence. Its fusion processing enables the further mining of data and utilization of data value and broadens the path for the sharing and dissemination of media data. However, it also faces serious problems in terms of protecting user and data privacy. Many privacy protection methods have been proposed to solve the problem of privacy leakage during the process of data sharing, but they suffer from two flaws: 1) the lack of algorithmic frameworks for specific scenarios such as dynamic datasets in the media domain; 2) the inability to solve the problem of the high computational complexity of ciphertext in multi-source data privacy protection, resulting in long encryption and decryption times. In this paper, we propose a multi-source data privacy protection method based on homomorphic encryption and blockchain technology, which solves the privacy protection problem of multi-source heterogeneous data in the dissemination of media and reduces ciphertext processing time. We deployed the proposed method on the Hyperledger platform for testing and compared it with the privacy protection schemes based on k-anonymity and differential privacy. The experimental results show that the key generation, encryption, and decryption times of the proposed method are lower than those in data privacy protection methods based on k-anonymity technology and differential privacy technology. This significantly reduces the processing time of multi-source data, which gives it potential for use in many applications.

24 February 2026 in Sensors
Verified
Journal Article
Open access
PDF file
JLJiafu Lv
Jiafu Lv
DZDi Zhang
Di Zhang
KHKe Han
Ke Han
QWQi Wu
Qi Wu
SCSanxing Cao
Sanxing Cao
Eye tracking in virtual reality (VR) head-mounted displays poses substantial engineering challenges, particularly under immersive display configurations with large fields of view (FOV), where optical layout, illumination, and image acquisition impose nontrivial system constraints. To address these design constraints, we present an integrated near-eye eye-tracking prototype tailored for immersive VR headsets, combining customized hardware components and a real-time software pipeline. The proposed system integrates optimized near-eye illumination and image acquisition with a pupil detection module and a deep learning-based gaze-vector estimation model, forming a real-time software pipeline for stable end-to-end gaze mapping under fixed calibration conditions. Under identical system settings, calibration procedures, and gaze-point mapping conditions, we evaluate the proposed gaze-vector estimation model through a controlled model-level ablation. The attention-enhanced model achieves an average angular deviation of 1.15°, corresponding to a 61.4% relative reduction compared with a baseline ResNet-152 model without attention. To demonstrate the usability of the system outputs at the application level, we further implement a real-time visualization example that integrates pupil diameter, gaze vectors, and blink events to depict the temporal evolution of eye-movement signals. This work provides a cost-effective and reproducible engineering reference for near-eye eye-movement acquisition and visualization in immersive VR settings and serves as a technical foundation for subsequent interaction design or behavioral analysis studies.

13 May 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
DWDong Wang
Dong Wang
SCSanxing Cao
Sanxing Cao
BTBaihui Tang
Baihui Tang
Speech-driven 3D cartoon facial animation remains underexplored due to the difficulty of handling heterogeneous geometries with exaggerated proportions, limited generalization to diverse unseen subjects, and the scarcity of datasets. To address these challenges, we propose Let Toon Talk, a two-stage cascaded framework that effectively mitigates these bottlenecks in both modeling and driving. It enables one-shot, speech-synchronized 3D animation from a single unseen humanoid cartoon image, driven by arbitrary audio. Specifically, for avatar modeling, we propose a parametric adaptation mechanism to capture diverse heterogeneous facial topologies, which subsequently guides a feed-forward reconstruction module to create high-quality 3D Gaussian Splatting (3DGS) avatars. Building upon this, for speech driving, we introduce an Identity-Adaptive Flow Matching network. This generative module effectively maps audio to precise facial dynamics, achieving identity-adaptive motion synthesis for diverse humanoid cartoon characters without per-subject pretraining. Furthermore, we construct a hybrid cartoon talking-face dataset with a systematic curation strategy to bridge the data gap. Extensive experiments demonstrate that our framework produces high-quality, temporally coherent animations, exhibiting effective generalization on unseen structurally humanoid cartoon characters.

18 June 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
BTBaihui Tang
Baihui Tang
SCSanxing Cao
Sanxing Cao
Face segmentation is a critical component in new media post-production, enabling the precise separation of facial regions from complex backgrounds at the pixel level. With the increasing demand for flexible and efficient segmentation solutions across diverse media scenarios—such as variety shows, period dramas, and other productions—there is a pressing need for adaptable methods that can perform reliably under varying conditions. However, existing approaches primarily depend on fully supervised learning, which requires extensive manual annotation and incurs high labor costs. To overcome these limitations, we propose a novel weakly supervised face segmentation framework that leverages large-scale vision models to automatically generate high-quality pseudo-labels. These pseudo-labels are then used to train segmentation networks in a dual-model architecture, where two complementary models collaboratively enhance segmentation performance. Our method significantly reduces the reliance on manual labeling while maintaining competitive accuracy. Extensive experiments demonstrate that our approach not only improves segmentation precision and efficiency but also streamlines post-production workflows, lowering human effort and accelerating project timelines. Furthermore, this framework reduced reliance on annotations in the field of weakly supervised learning for facial image processing in the new media post-production scenario.