- Advanced Sensor and Control Systems
- Neural Networks and Applications
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Speech Recognition and Synthesis
- Phonetics and Phonology Research
- Face and Expression Recognition
- Data Mining Algorithms and Applications
- Handwritten Text Recognition Techniques
- Emotion and Mood Recognition
- Topic Modeling
- Advanced Computational Techniques and Applications
- Multimodal Machine Learning Applications
- Integrated Circuits and Semiconductor Failure Analysis
- Video Surveillance and Tracking Methods
- Image and Signal Denoising Methods
- Medical Image Segmentation Techniques
- Fuzzy Logic and Control Systems
- Data Management and Algorithms
- Speech and Audio Processing
- Domain Adaptation and Few-Shot Learning
- Reading and Literacy Development
- Natural Language Processing Techniques
- Model Reduction and Neural Networks
- Advanced Algorithms and Applications
East China Normal University
2021-2024
Northwestern Polytechnical University
2020-2024
Shanxi Provincial Children's Hospital
2024
Central University of Finance and Economics
2024
Chinese Academy of Sciences
2003-2023
Anhui University
2023
Hefei Institutes of Physical Science
2023
Institute of Intelligent Machines
2023
Louisiana State University
2020-2022
Faculty (United Kingdom)
2022
We propose a new matching-based framework for semi-supervised video object segmentation (VOS). Recently, state-of-the-art VOS performance has been achieved by algorithms, in which feature banks are created to store features region matching and classification. However, how effectively organize information the continuously growing bank remains under-explored, this leads inefficient design of bank. introduce an adaptive update scheme dynamically absorb discard obsolete features. also confidence...
This article presents a new text-to-image (T2I) generation model, named distribution regularization generative adversarial network (DR-GAN), to generate images from text descriptions improved learning. In DR-GAN, we introduce two novel modules: semantic disentangling module (SDM) and normalization (DNM). SDM combines the spatial self-attention mechanism (SSAM) loss (SDL) help generator distill key information for image generation. DNM uses variational auto-encoder (VAE) normalize denoise...
In order to solve the problems of single movement pattern recognition information and low accuracy multi-joint upper limb exoskeleton rehabilitation training, a multimodal fusion method with human surface electromyography(sEMG) electrocardiogram(ECG) was proposed, an Inception-Sim model for motion designed. Integrating advantages information, inspired by convolutional neural network processing image classification problem, original signal converted into Gramian Angular Summation/Difference...
Vector representations for language have been shown to be useful in a number of Natural Language Processing tasks. In this paper, we aim investigate the effectiveness word vector problem Sentiment Analysis. particular, target three sub-tasks namely sentiment words extraction, polarity detection, and text prediction. We over different data evaluate quality domain-dependent vectors. has used compute various vector-based features conduct systematically experiments demonstrate their...
In this paper, we propose a new technique to achieve accurate decomposition of process variation by efficiently performing spatial pattern analysis.We demonstrate that the spatially correlated systematic can be accurately represented linear combination small number "templates".Based upon observation, an efficient sparse regression algorithm is developed extract most adequate templates represent variation.In addition, robust proposed automatically remove measurement outliers.We further...
Readers' eye movements were recorded to examine the role of character positional frequency on Chinese lexical acquisition during reading and its possible modulation by word spacing. In Experiment 1, three types pseudowords constructed based each character's frequency, providing congruent, incongruent, no segmentation information. Each pseudoword was embedded into two sets sentences, for learning test phases. phase, half participants read sentences in word-spaced format, unspaced format. all...
Linear registration is often the crucial first step for various types of image analysis. Although this mathematically simple, failure not uncommon. When investigating brain by magnetic resonance imaging (MRI), target organ but existence other tissues, in addition to a variety fields view, different locations, orientations and anatomical features, poses some serious fundamental challenges. Consequently, number algorithms have been put forward minimize potential errors. In present study, we...
Word spacing is important in guiding eye movements during spaced alphabetic reading. Chinese unspaced and it remains unclear as to how readers segment identify words We conducted two parallel experiments investigate whether the positional probabilities of initial final characters a multicharacter word affected segmentation identification Two-character were selected targets. In Experiment 1, character's probability was manipulated being either high or low, character kept identical across...
To address the problem of approximation and prediction complex time-varying system, this paper proposes a parallel process neural networks predication method based on general models. Firstly, whole is divided into several small time intervals; then, are constructed respectively in intervals to disperse load networks. According theory orthogonal function basis expansion functional space, learning algorithm above model deduced; finally, results series for sunspots shows that proposed can...
In this paper, we study the vocabulary design problem in Uyghur large continuous speech recognition (LVCSR). is an agglutinative language which words can be formed by concatenating several suffixes to stem. As a result, number of word types unlimited. If used as unit, out-of-vocabulary (OOV) rate will very with typical sizes 60 k-100 k. To avoid problem, split into stems and use these sub-words units. Speech experiments are performed two test sets, one including sentences books another...
Title: Determining the Optimal Number of Clusters by an Extended RPCL Algorithm | Keywords: Clustering, algorithm, Competitive learning, Elliptical basis function networks, Speaker verification Author: Xin Li, Man Wai Mak and Chi Kwong Li