A5062971861- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Neural dynamics and brain function
- Video Analysis and Summarization
- Music and Audio Processing
- Neuroscience and Neural Engineering
- Human Pose and Action Recognition
- EEG and Brain-Computer Interfaces
- Neural Networks and Applications
- Data Management and Algorithms
- Multimodal Machine Learning Applications
- Reinforcement Learning in Robotics
- Neurobiology and Insect Physiology Research
- Speech and Audio Processing
- Video Surveillance and Tracking Methods
- Advanced Steganography and Watermarking Techniques
- Higher Education and Teaching Methods
- Ergonomics and Musculoskeletal Disorders
- Human Motion and Animation
- Color perception and design
- Image Processing Techniques and Applications
- Remote-Sensing Image Classification
- Animal Vocal Communication and Behavior
- Engineering Education and Curriculum Development
- Advanced Memory and Neural Computing
Zhengzhou University
2019-2025
Communication University of China
2013-2024
Guangdong Open University
2024
Guangxi Zhuang Autonomous Region Brain Hospital
2020-2024
Beijing Foreign Studies University
2023
University of Science and Technology Beijing
2022
Harbin Institute of Technology
2006-2021
Zhengzhou University of Industrial Technology
2018-2020
ITRI International
2019
Huaqiao University
2007-2015
Learning representations for multimedia content is critical recommendation. Current representation learning methods roughly fall into two groups: (1) using the historical interactions to create ID embeddings of users and items, (2) treating multi-modal data as side information items enrich their embeddings. Each user-item interaction offers supervisory signal optimize by traditional supervised paradigm. Due overlook patterns (<inline-formula><tex-math...
Transformer typically enjoys larger model capacity but higher computational loads than convolutional neural network (CNN) in vision tasks. In this letter, the advantages of such two networks are fused for achieving effective and efficient real image denoising. We propose a hybrid denoising based on Encoder Convolutional Decoder Network (TECDNet). The novel radial basis function (RBF) attention is used as encoder to improve representation capability overall model. decoder, residual CNN...
Learning engagement is a crucial predictor of academic achievement. It essential to understand the factors influencing learning among nursing students, especially from learner's perspective, which notably scarce but vital for designing effective educational interventions.
Grain size is one of the most important parameters for metallographic microstructure analysis, which can partly determine material performance. The measurement grain based on accurate image segmentation methods, include traditional processing methods and emerging machine-learning-based methods. Unfortunately, hardly segment grains correctly from images with low contrast blurry boundaries. Moreover, proposed need a large dataset to train model deal challenge complex fuzzy boundaries...
With the development of artificial intelligence, multiagent algorithms have been applied to many real-time strategy games. Making plans on human being is gradually passing away, especially in combat scenarios. Cognitive electronic warfare (CEW) a complex and challenging work due sensitivity data sources. There are few studies CEW. In past, wargame simulations depended differential equations war theory, which resulted high time resource costs. future, as other intelligence theories developed,...
Drug-resistant epilepsy (DRE) patients typically require surgical intervention or neurostimulation. Therefore, accurate localization of the seizure onset zone (SOZ) is essential for effective clinical intervention. Although some physiologically meaningful parameters neural computational models show substantial differences across brain regions during seizures, few studies pay attention to applying these model SOZ localization. To investigate whether parameter can be used localization,...
As eating-out became an indispensable part of our daily lives, demand for the food recognition unfamiliar restaurant increased significantly due to health-care. Although there are many researches on generic recognition, relatively fewer studies image recognition. Meanwhile, it becomes extremely challenging insufficient image. Prototypical network is common utilized address such a task in recent years. methods based prototypical achieve impressive results capturing similarities feature same...
In general, dance is always associated with music to improve stage performance effect. As we know, artificial arrangement consumes a lot of time and manpower. While automatic based on input video perfectly solves this problem. the cross-modal generation task, take advantage complementary information between two modalities facial expressions movements. Then present Dance2MusicNet (D2MNet), an autoregressive model dilated convolution, which adopts feature vectors, style beats, as control...
The main idea of multimedia recommendation is to introduce the profile content documents as an auxiliary, so endow recommenders with generalization ability and gain better performance. However, recent studies using non-uniform datasets roughly fuse single-modal features into multi-modal adopt strategy directly maximizing likelihood user preference scores, leading bias. Owing defect in architecture, there still room for improvement recommendation.
Car License Plate Recognition (CLPR) is the key technology of achievement intelligent transport system, and vehicle license plate location plays a very important part in whole system. The article presents an effective method based on wavelet transform to locate plate. Firstly, general profile gained by extracting HxLy subband image after taking (vertical texture). Then, vertical will be smeared horizontally. Finally, results are obtained 8-connected regional growth filtering out non-text...
This article has been retracted. A retraction notice can be found at https://doi.org/10.3233/JIFS-219433.
Histograms of Oriented Gradients (HOG) feature has been successfully used in pedestrian detection and achieves high accuracy. This paper introduces a content retrieval algorithm based on improved HOG. The method two steps which are adjusting the HOG structure by scanning image with sliding window reducing dimension principle component analysis (PCA) technique. experimental results show that: precision rate this significantly compared transforming size images to calculate extracting color feature.
Front Vehicle Detection is the key and difficult point of technology research for intelligent vehicle. In this paper digital image firstly binarized through enhancement, threshold segmentation noise eliminating along with recent patents described. Then hypothesis generation done according to structure, shape, aspect ratio vehicle shadow at bottom On basis, features extraction performed a Gabor method based on improved weightings selected samples background samples. The extracted vector...
Goal-directed spatial learning is crucial for the survival of animals, in which formation route from current location to goal one central problems. A distributed brain network comprising hippocampus and prefrontal cortex has been shown support such capacity, yet it not fully understood how most similar regions birds, (Hp) nidopallium caudolaterale (NCL), cooperate during goal-directed learning. Hence, we examined neural activity Hp-NCL pigeons explored connectivity dynamics a task. We found...
Abstract A pigeon robot is an ideal experimental animal for research in flying robots. The majority of current publications have entailed electrical stimulation the motor nuclei to regulate movement forcibly, and although a “virtual fear” behavior model has been proposed, structure, location, function that generate fear emotions remain obscure. Previous studies shown Stratum Griseum Periventriculare (SGP) pigeons homologous mammalian periaqueductal gray (PAG), which plays essential role...
Currently, micro-videos have grown explosively on various online social platforms. Accordingly, how to encode them yield effective representation attracts our attention. NeXtVLAD is such an network that aggregates frame-level features into a compact supervector. However, the discriminant capability of supervector still limited due lack non-linear transformation and L2 normalization at head tail original network, respectively. In order address problems, we propose improved neural...