A5019371593- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Gait Recognition and Analysis
- Internet Traffic Analysis and Secure E-voting
- Privacy-Preserving Technologies in Data
- Cryptography and Data Security
- Advanced Vision and Imaging
- User Authentication and Security Systems
- EEG and Brain-Computer Interfaces
- Retinal Imaging and Analysis
- Deception detection and forensic psychology
- Context-Aware Activity Recognition Systems
- Adversarial Robustness in Machine Learning
- ECG Monitoring and Analysis
- Retinal Diseases and Treatments
- Generative Adversarial Networks and Image Synthesis
- Phonocardiography and Auscultation Techniques
- Network Security and Intrusion Detection
- Digital Media Forensic Detection
- Advanced Malware Detection Techniques
- Biometric Identification and Security
- Glaucoma and retinal disorders
University of Electronic Science and Technology of China
2019-2025
Privacy in the Internet of Things is a fundamental challenge for Ubiquitous healthcare systems that depend on data aggregated and collaborative deep learning among different parties. This paper proposes MSCryptoNet, novel framework enables scalable execution conversion state-of-the-art learned neural network to MSCryptoNet models privacy-preservation setting. We also design method approximation activation function basically used convolutional (i.e., Sigmoid Rectified linear unit) with low...
With the rapid development of Mobile Internet and Industrial Things, a variety applications put forward an urgent demand for user device identity recognition. Digital with hidden characteristics is essential both individual users physical devices. assistance multimodalities as well fusion strategies, recognition can be more reliable robust. In this survey, we turn to investigate concepts limitations unimodal recognition, motivation, advantages multimodal summarize technologies via feature...
Abstract Human activity recognition (HAR) generates a massive amount of the dataset from Internet Things (IoT) devices, to enable multiple data providers jointly produce predictive models for medical diagnosis. That accuracy is greatly improved when trained on large number datasets these untrusted cloud server very significant and raises privacy concerns. With migration deep neural network (DNN) in learning experience HAR, we present privacy-preserving DNN model known as Multi-Scheme...
Human skeleton data, which has served in the aspect of human activity recognition, ought to be most representative biometric characteristics due its intuitivity and visuality. The state-of-the-art approaches mainly focus on improving modeling spatial correlations within graph topologies. However, interframes motional representations are also vital importance, we argue that they worth paying attention exploring. Therefore, a temporal refinement module with contrastive learning mechanism is...
The wide application of human action recognition in the field computer vision makes it a hot research topic past decades. In recent years, prevalence deep sensors and proposal real-time skeleton estimation algorithm based on images make sequence attract increasing attention researchers. Most existing work is aimed at extracting spatial information different joint nodes frame, but they do not fully consider combination temporal features. At same time, joints were regarded as equally...
The electrocardiogram diagnosis played an important role in early arrhythmia prevention and cardiovascular disease detection. How to analysis detect the automatically becomes a challenging task clinical practice. Convolutional neural network based approaches have been widely applied field of automatic detection analysis. However, existing methods fails preserve abundant essential latent representatives from multiple dimensions. To address these problems, dual spatial-channel aggregation...
Personal information is a kind of sensitive data that users prefer keeping in private. However, with the development Mobile Internet, various novel applications generate large amount data, which could be used to correlate and further infer their private information. This paper argues smartphone application usage would incur potential privacy issue, proposes scheme verify case study gender leaks while built-in accelerator used. In particular, an Android OS was firstly developed collect...