A5100437809- Topic Modeling
- Radio Frequency Integrated Circuit Design
- Anomaly Detection Techniques and Applications
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Natural Language Processing Techniques
- Advanced Algorithms and Applications
- Advancements in PLL and VCO Technologies
- Human Pose and Action Recognition
- Speech and Audio Processing
- VLSI and Analog Circuit Testing
- Indoor and Outdoor Localization Technologies
- Advanced Image and Video Retrieval Techniques
- Advanced Sensor and Control Systems
- Energy Load and Power Forecasting
- Speech Recognition and Synthesis
- Blind Source Separation Techniques
- Video Surveillance and Tracking Methods
- Analog and Mixed-Signal Circuit Design
- Advanced Wireless Communication Techniques
- Advanced Battery Technologies Research
- Neural Networks and Applications
- Microwave Engineering and Waveguides
- Embedded Systems and FPGA Design
- Industrial Vision Systems and Defect Detection
Nanjing Drum Tower Hospital
2025
Technical University of Munich
2025
China Academy of Information and Communications Technology
2022-2025
Hebei Agricultural University
2014-2024
Southeast University
2015-2024
Chinese Academy of Sciences
2009-2024
PLA Army Engineering University
2024
National University of Defense Technology
2024
Army Medical University
2024
Daping Hospital
2024
Adversarial attacks are carried out to reveal the vulnerability of deep neural networks. Textual adversarial attacking is challenging because text discrete and a small perturbation can bring significant change original input. Word-level attacking, which be regarded as combinatorial optimization problem, well-studied class textual attack methods. However, existing word-level models far from perfect, largely unsuitable search space reduction methods inefficient algorithms employed. In this...
Using solar photovoltaics (PV) to help a microgrid (MG) operator for cost reduction may not be straightforward problem due the intermittent nature of PV power generation and unpredictable load demands. One potential way address this challenge is use batteries that can store surplus energy whenever possible supply back MG when needed. In context, paper proposes battery management system (BEMS) an MG, in which PVs diesel generators (DGs) are primary sources electricity. The novelty proposed...
Recent convolutional neural networks (CNNs)-based image processing methods have proven that CNNs are good at extracting features of spatial data. In this letter, we present a CNN-based modulation recognition framework for the detection radio signals in communication systems. Since frequency variation with time is most important distinction among different types, transform 1-D into spectrogram images using short-time discrete Fourier transform. Furthermore, analyze statistical and use...
Accurate short-term photovoltaic (PV) power forecasting is of great significance for the safe and stable operation system. Spatial information from neighboring PV sites contributes to improving performance. However, most current methods considering spatial indiscriminately use all data modeling, which will lead redundancy, resulting in low accuracy. Therefore, this paper proposes a solar method based on optimal graph structure surrounding spatio-temporal correlations. Firstly, analyzed...
In this paper, we present an automatic modulation recognition framework for the detection of radio signals in a communication system. The considers both deep convolutional neural network (CNN) and long short term memory network. Further, propose pre-processing signal representation that combines in-phase, quadrature fourth-order statistics modulated signals. presented data allows our CNN LSTM models to achieve 8% improvements on testing dataset. We compare accuracy proposed methods with...
Discovering social relations, e.g., kinship, friendship, etc., from visual contents can make machines better interpret the behaviors and emotions of human beings. Existing studies mainly focus on recognizing relations still images while neglecting another important media--video. On one hand, actions storylines in videos provide more cues for relation recognition. other key persons may appear at arbitrary spatial-temporal locations, even not same image beginning to end. To overcome these...
There has been an increasing demand for infrastructureless localization. Current approaches involving inertial measurement unit (IMU) generally utilize step detection and counting to estimate the displacement. However, accuracy is affected, because sizes are neglected. Some groups have proposed algorithms that involve placing IMU on foot size, but users commented it affects their walking. Hence, this paper presents a new method both forward displacement orientation with placed at upper...
Abstract With the improvement of electronic circuit production methods, such as reduction component size and increase density, risk defects is increasing in line. Many techniques have been incorporated to check for failed solder joints, X-ray imaging, optical imaging thermal among which can inspect external internal defects. However, some advanced algorithms are not accurate enough meet requirements quality control. A lot manual inspection required that increases specialist workload. In...
Most of the research to date focuses on bilingual sign language translation (BSLT). However, such models are in-efficient in building multilingual systems. To solve this problem, we introduce multilin-gual (MSLT) task. It aims use a single model complete between multiple languages and spoken languages. Then, propose MLSLT, first MSLT model, which contains two novel dy-namic routing mechanisms for controlling degree ofpa-rameter sharing different Intra-layer language-specific controls...
This paper focuses on fine-grained human parsing in images. is a very challenging task due to the diverse person appearance, semantic ambiguity of different body parts and clothing, extremely small targets. Although existing approaches can achieve significant improvement by pyramid feature learning, multi-level supervision, joint learning with pose estimation, still far from being solved. Different approaches, we propose Braiding Network, named as BraidNet, learn complementary semantics...
Along with the popularization of small unmanned aerial vehicles (UAVs), societal concerns related to security, privacy, and public safety have gained more attention, thus opening a new avenue for UAV surveillance. However, conventional radar technologies pose challenges surveillance UAVs due high cost, cross section, low flying altitude, slow speed. In this article, we propose novel micro-Doppler signature-based method using machine learning techniques, detection, classification,...
Discovering social relations in images can make machines better interpret the behavior of human beings. However, automatically recognizing is a challenging task due to significant gap between domains visual content and relation. Existing studies separately process various features such as faces expressions, body appearance, contextual objects, thus they cannot comprehensively capture multi-granularity semantics, scenes, regional cues persons, interactions among persons objects. To bridge...