A5076188308- Face and Expression Recognition
- Face recognition and analysis
- Video Surveillance and Tracking Methods
- Advanced Algorithms and Applications
- Emotion and Mood Recognition
- Image and Video Stabilization
- Advanced Computing and Algorithms
- Industrial Vision Systems and Defect Detection
- Prosthetics and Rehabilitation Robotics
- Stroke Rehabilitation and Recovery
- Vehicle Noise and Vibration Control
- Robotic Locomotion and Control
- Vehicle Dynamics and Control Systems
- Video Analysis and Summarization
- Human Pose and Action Recognition
- Image and Object Detection Techniques
- Muscle activation and electromyography studies
- Welding Techniques and Residual Stresses
- EEG and Brain-Computer Interfaces
- Neural Networks and Applications
- Gait Recognition and Analysis
- IoT and GPS-based Vehicle Safety Systems
- Music and Audio Processing
- Vehicular Ad Hoc Networks (VANETs)
- Machine Fault Diagnosis Techniques
Changchun University of Technology
2012-2025
IRD Fuel Cells (Denmark)
2024
Changchun University
2014
Jilin University
2010-2012
This paper proposes a fractional order optimization method of multi-variable Grey model (GM( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$r$</tex-math> </inline-formula> ,2)) based on full transfer learning LSTM network. Firstly, GM( ,2) was built with adhesion coefficient as input variable, namely, correlation factor sequence, and driving intention output system behaviour characteristic sequence....
In order to solve the problem which key information of facial expression missed under partial occlusion, this paper proposes a recognition method based on Weber Local Descriptor (WLD) histogram feature and decision fusion. Firstly, image is divided into several non-overlapping rectangle regions with equal size WLD used extract features each region. The purpose that face spatial be ready for subsequent Secondly, region further blocks sub-block computed combined as so discriminative accurately...
This paper describes a method for recognition of continuous facial expression change in video sequences. ASM automatically localizes the feature points first frame and then tracks through frames. After that step is selection 20 optimal key points, those which most with changes expression. Since distance geometric features, set displacement vectors, high dimensions, it mapped into low dimensional space, called by applying PCA expansion. Then estimation input image achieved projecting on to...
Abstract Reconstruction‐based methods are commonly used in industrial visual anomaly detection. They rely on a well‐reconstructed normal mode of the model. However, it is difficult to manage boundary generalization. The strength model's generalization capability can directly affect fidelity reconstruction, resulting occurrence false positives. To address above challenges, novel dual branch reconstruction detection approach proposed control model at two dimensions. It reconstructs abnormal...
Most unsupervised methods of person re-identification (Re-ID) obtain pseudo-labels through clustering. However, in the process clustering, hard quantization loss caused by clustering errors will make model produce false pseudo-labels. In order to solve this problem, an based on softened labels training method is proposed. The innovation that correlation among image features used find reliable positive samples and train them a smooth manner. To further explore features, some modules are...
Recognizing occluded facial expressions in the wild poses a significant challenge. However, most previous approaches rely solely on either global or local feature-based methods, leading to loss of relevant expression features. To address these issues, feature fusion residual attention network (FFRA-Net) is proposed. FFRA-Net consists multi-scale module, and module. The module divides intermediate map into several sub-feature maps an equal manner along channel dimension. Then, convolution...
The problem in parameter selection of least squares support vector machine (LS-SVM) restricts the development LS-SVM, In order to choose optimal parameters LS-SVM automatically, we proposed an improved particle swarm optimization (PSO) algorithm which can not only increase convergent speed but also improve overall searching ability algorithm. PSO increases avoiding local optimum effectively. We use automatically facial expression recognition system. experimental results show that method with...
Rehabilitation training is usually a long-term systematic process for the patients of lower limb paralysis. Using robot-assisted rehabilitation can improve efficiency, reduce costs and burden on physical therapists, therefore, resources be saved. For extremity with motor dysfunction, new three degree freedom (3DOF) robot (LLRR) designed, investigated analyzed which used in passive phase patient training. At first, simple LLRR structure designed that adjusted to fit at hip, knee, ankle...
A normalization method of road adhesion coefficient and tire cornering stiffness is proposed to provide the significant information for vehicle direct yaw-moment control (DYC) system design. This carried out based on a fractional-order multi-variable gray model (FOMVGM) long short-term memory (LSTM) network. FOMVGM used generate training data testing LSTM network, network employed predict with coefficient. In addition that, represented by can be built lateral dynamic participate in DYC...
Learning discriminative features for facial expression recognition (FER) in the wild is a challenging task due to significant intra-class variations, inter-class similarities, and extreme class imbalances. In order solve these issues, contrastive-learning-based extra-contrast affinity network (ECAN) method proposed. The ECAN consists of feature processing two proposed loss functions, namely extra negative supervised contrastive (ENSC loss) multi-view (MVA loss). provides current historical...
Most existing methods tackle the problem of occluded person re-identification (ReID) by utilizing auxiliary models, resulting in a complicated and inefficient ReID framework that is unacceptable for real-time applications. In this work, speed-up named SUReID proposed to mitigate occlusion interference while speeding up inference. The consists three key components: hierarchical token sparsification (HTS) strategy, non-parametric feature alignment knowledge distillation (NPKD), noise data...
Current multimodal fusion models can only accept a speci c number of modal ities and cannot change the input modalities when are missing or added. The overall structure needs to be modi ed in order re data, scalability is poor. Based on this, this paper proposes progressive network that attached pre-trained model. It fuses data independently pre-training model trains jointly with it. By fusing collects relevant information about task fed into focuses each modality selects one as main fused...
Facial Expression Recognition (FER) is a challenging task in computer vision that aims to infer people's emotional states by analyzing and recognizing their facial expressions. However, the extraction of expression features from images usually interfered other (e.g. pose appearance), which makes it difficult improve performance model. To further optimize model performance, an approach called Multi-level Integration Disentangled Generative Adversarial Network (MID-GAN) introduced address...
Increasing the linear modulation range
Facial expression representation based on Gabor features has attracted more and attention achieved great success in facial recognition for some favorable attributes of wavelets such as spatial locality orientation selectivity. A large number are produced with varying parameters the position, scale filters, which cause huge computational complexity. In existing methods, useful discriminatory information may be lost due to down sampling directly. To reduce loss, a method block statistics is...
The problem in accurate and fast face detection tracking video images with complex background is studied this paper. In order to meet both of speed accuracy, a method integrating the skin color segmentation Split up Sparse Network Winnows (SNoW) presented. This kind can not only utilize advantage remove interference background, but also has SNoW accurately classifying feathers. More importantly, inherit their common speed. Based on these advantages, used detect human areas then use way...