A5082286929- Speech and Audio Processing
- Speech Recognition and Synthesis
- Music and Audio Processing
- Phonetics and Phonology Research
- Digital Media Forensic Detection
- Advanced Steganography and Watermarking Techniques
- Semantic Web and Ontologies
- Advanced Graph Neural Networks
- Natural Language Processing Techniques
- AI in cancer detection
- Multi-Agent Systems and Negotiation
- Advanced Image Processing Techniques
- Brain Tumor Detection and Classification
- Anomaly Detection Techniques and Applications
- Video Surveillance and Tracking Methods
- Medical Image Segmentation Techniques
- Lattice Boltzmann Simulation Studies
- Fluid Dynamics Simulations and Interactions
- Biomedical Text Mining and Ontologies
- Energy Load and Power Forecasting
- Video Analysis and Summarization
- Advanced Image Fusion Techniques
- Data Visualization and Analytics
- Medical Imaging and Analysis
- Voice and Speech Disorders
Tianjin University
2016-2025
Institute of Information Engineering
2016
Japan Advanced Institute of Science and Technology
2005-2010
With a focus on abnormal events contained within untrimmed videos, there is increasing interest among researchers in video anomaly detection. Among different detection scenarios, weakly-supervised poses significant challenge as it lacks frame-wise labels during the training stage, only relying video-level coarse supervision. Previous methods have made attempts to either learn discriminative features an end-to-end manner or employ two-stage self-training strategy generate snippet-level pseudo...
The existing learning resource recommendation systems suffer from data sparsity and missing labels, leading to the insufficient mining of correlation between users courses. To address these issues, we propose a method based on graph contrastive learning, which uses construct an auxiliary task combined with main task, achieving joint resources. Firstly, interaction bipartite user course is input into lightweight convolutional network, embedded representation each node in obtained after...
While deep learning techniques, such as Convolutional neural networks (CNNs), show significant potential in medical applications, real-time detection of parathyroid glands (PGs) during complex surgeries remains insufficiently explored, posing challenges for surgical accuracy and outcomes. Previous studies highlight the importance leveraging prior knowledge, shape, feature extraction tasks. However, they fail to address critical multi-scale variability PG objects, resulting suboptimal...
Extended reality (XR) is a general term for virtual (VR), augmented (AR), and mixed (MR). By converting abstract digital expressions into intelligent feedback through figures, one can effectively compensate the poor performance of traditional learning in deep cognitive processing operational skills training. However, extant results are uncertain, only limited number studies have investigated influence mechanism heterogeneity among VR, AR, MR on procedural knowledge learning, higher-level...
Due to the fast transmission speed and severe health damage, COVID-19 has attracted global attention. Early diagnosis isolation are effective imperative strategies for epidemic prevention control. Most diagnostic methods is based on nucleic acid testing (NAT), which expensive time-consuming. To build an efficient valid alternative of NAT, this article investigates feasibility employing computed tomography images lungs as signals. Unlike normal lungs, parts infected with developed lesions,...
Community detection is an important task in social network analysis. Existing methods typically use the topological information alone, and ignore rich available content data. Recently, some researchers have noticed that user profiles can also benefit to community detection, hence combination of topology node contents has become a new hot topic. Some using both been proposed. However, they often suffer from two drawbacks: 1) cannot extract potential deep representation network; 2)...
Single image dehazing has always been a challenging problem in the field of computer vision. Traditional defogging methods use manual features. With development artificial intelligence, method based on deep learning developed rapidly. In this paper, we propose novel approach called NIN-DehazeNet for single image. This estimates transmission map by combining Network-in-Network with MSCNN(Single Image Dehazing via Multi-Scale Convolutional Neural Networks). test stage, estimate input hazy...
The generation-based data augmentation method can overcome the challenge caused by imbalance of medical image to a certain extent. However, most current research focus on images with unified structure which are easy learn. What is different that ultrasound structurally inadequate, making it difficult for be captured generative network, resulting in generated lacks structural legitimacy. Therefore, Progressive Generative Adversarial Method Structurally Inadequate Medical Image Data...
Automatic speaker verification (ASV) systems, which determine whether two speeches are from the same speaker, mainly focus on accuracy while ignoring inference speed. However, in real applications, both speed and essential. This study proposes cross-sequential re-parameterization (CS-Rep), a novel topology strategy for multi-type networks, to increase of models. CS-Rep solves problem that existing methods not suitable typical ASV backbones. When model applies CS-Rep, training-period network...
In this paper, a novel imperceptible, fragile and blind watermark scheme is proposed for speech tampering detection self-recovery. The embedded data content recovery calculated from the original discrete cosine transform (DCT) coefficients of host speech. information shared in frames-group instead stored one frame. trades off between waste problem coincidence problem. When part watermarked signal tampered with, can accurately localize area, area without any modification still be extracted....
Digital watermarking is an effective covert communication technique in the military field, which can solve problems traditional encryption. In general, Spread Spectrum (SS) algorithm well satisfy requirements of concealment and anti-interference process. However, it has a natural defect resisting desynchronization attacks. this paper, we propose Robust Feature Points Scheme (RFPS) against attacks by calculating maximum response value second-order derivative original audio signal. The will be...
The objective of the study is to develop a framework for automatic breast cancer detection with merging four imaging modes. Attempts were made tumor classification and segmentation; using multi-parametric Magnetic Resonance Imaging (MRI) method on tumors. MRI data obtained from 67 subjects 1.5T-MRI scanner. Four modes: T1 weighted, T2 Diffusion Weighted eTHRIVE sequences, dynamic-contrast-enhanced(DCE)-MRI parameters are acquired. proposed four-mode linkage backbone in classification, which...