Micro-expression recognition from local facial regions

MiE is a facial involuntary reaction that reflects the real emotion and thoughts of a human being. It is very difficult for a normal human to detect a Micro-Expression (MiE), since it is a very fast and local face reaction with low intensity. As a consequence, it is a challenging task for researchers to build an automatic system for MiE recognition. Previous works for MiE recognition have attempted to use the whole face, yet a facial MiE appears in a small region of the face, which makes the extraction of relevant features a hard task. In this paper, we propose a novel deep learning approach that leverages the locality aspect of MiEs by learning spatio-temporal features from local facial regions using a composite architecture of Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM). The proposed solution succeeds to extract relevant local features for MiEs recognition. Experimental results on benchmark datasets demonstrate the highest recognition accuracy of our solution with respect to state-of-the-art methods.     

Flash cosintering of a lanthanum strontium cobalt ferrite nanofibre/Gd-doped ceria bilayer structure

For solid oxide fuel cells, an important structural requirement is that the electrolyte layer needs to be dense and the electrode layer porous, which is difficult to obtain by conventional cosintering. In this work, flash cosintering of a double layer structure consisting of a Gd-doped ceria substrate with a lanthanum strontium cobalt ferrite nanofibre coating is investigated. Experimental and finite element modelling results reveal that when the LSCF layer is connected to the electrode, the heat is concentrated in the LSCF layer, which leads to a huge temperature gradient and introduces severe cracking. When the LSCF layer is electrically isolated from the electrode, the heat is concentrated in the GDC layer, and the temperature gradient is dramatically reduced. In this situation, the density of GDC can reach 92.86% while a high porosity of 52.26% is maintained in the LSCF layer, which is higher than that of the conventional cosintered sample.     

Sealing behaviour of glass-based composites for oxygen transport membranes

In this study, seven different filler materials in different proportions were added to a Ba-Ca-Si glass matrix “H” to investigate new sealant with higher thermal expansion coefficient (CTE) value and good sealing performance for application in oxygen transport membrane (OTM). SrTi_0.75Fe_0.25O_3-δ (STF25) was used as an OTM, and the sealing partners were ferritic steel Aluchrom and pre-oxidized Aluchrom. Compatibility tests were carried out to investigate the feasibility of the composites. Higher CTE values were found in dilatometer tests on composite samples by adding 40 wt% Ag (HAg40) and 30 wt% Ni-Cr (HNC30). Gas-tightness measurements of sandwiched samples produced appropriate helium leakage rates in the range of 10^?6 mbar·l·s^?1. Sealing behaviour of sealants HAg40 and HNC30 were investigated by joining STF25 and as-delivered/pre-oxidized Aluchrom together. Scanning electron microscopy (SEM) on cross-sections of the joints revealed a homogeneous microstructure and good adherence of the glass sealants to support metals and STF25.     

基于PU学习的链接预测方法

基于分类的链接预测方法中,由于链接未知节点对的大规模性与不确定性,选择可靠负例成为构造链接预测分类器的难点问题.为此,文中提出基于正例和无标识样本(PU)学习的链接预测方法.首先,提取节点对的拓扑信息以构造样本集.再利用社区结构确定候选负例的分布,基于分布进行多次欠采样,获得多个候选负例子集,集成多个负例集与正例集中构建的分类器选择可靠负例.最后基于正例与可靠负例构造链接预测分类器.在4个网络数据集上的实验表明文中方法预测结果较优.     

Hierarchical Attention for Part-Aware Face Detection

International Journal of Computer Vision - Expressive representations for characterizing face appearances are essential for accurate face detection. Due to different poses, scales, illumination,...     

Two-dimensional temperature distribution estimation for a cross-flow planar solid oxide fuel cell stack

The uniform temperature distribution of a cross-flow planar solid oxide fuel cell (SOFC) stack plays an essential role in stack thermal safety and electrical property. However, because of the strict requirements in stack sealing struture, it is hard to acquire the temperature inside the stack using thermal detection devices within an acceptable cost. Therefore, accurately estimating the two-dimensional (2-D) temperature distribution of the cross-flow stack is crucial for its thermal management. In this paper, Firstly, a 2-D mechanism model of a cross-flow planar SOFC stack is established. The stack is divided into 5*5 nodes along the gas flow directions, which can reflect the stack states with moderate computational burden. Then, experimental test data is utilized to modify and validate the stack model, guaranteeing the model accuracy as well as the reliability of model-based state estimator design. Finally, easily-measured stack inputs and outputs are selected, and a temperature distribution estimator combined with unscented kalman filter (UFK) approach is developed to achieve accurate and fast temperature distribution estimation of a cross-flow SOFC stack. Simulation results demonstrate that the UKF-based temperature distribution estimator can precisely and quickly achieve the temperature distribution estimation of the cross-flow stack under both static state and dynamic state changes and is applicable to cross-flow stacks with different size or cell number as well, the maximum estimated absolute error is less than 0.15 K with an absolute error rate of 0.015%, which indicates the developed estimator has good estimation performances.     

Affine hull based target representation for visual tracking

Handling appearance variations is a challenging issue in visual tracking. Existing appearance models are usually built upon a linear combination of templates. With such kind of representation, accurate visual tracking is not desirable when heavy appearance variations are in presence. Under the framework of particle filtering, we propose a novel target representation for tracking. Namely, the target candidates are represented by affine combinations of a template set, which leads to better capability in describing unseen target appearances. Additionally, in order to adapt this representation to dynamic contexts across a video sequence, a novel template update scheme is presented. Different from conventional approaches, the scheme considers both the importance of one template to a target candidate in the current frame and the recentness of the template that is kept in the template set. Comprehensive experiments show that the proposed algorithm achieves superior performances in comparison with state-of-the-art works.     

Constrained differential evolution with multiobjective sorting mutation operators for constrained optimization

Differential evolution (DE) is a simple and powerful evolutionary algorithm for global optimization. DE with constraint handling techniques, named constrained differential evolution (CDE), can be used to solve constrained optimization problems (COPs). In existing CDEs, the parents are randomly selected from the current population to produce trial vectors. However, individuals with fitness and diversity information should have more chances to be selected. This study proposes a new CDE framework that uses nondominated sorting mutation operator based on fitness and diversity information, named MS-CDE. In MS-CDE, firstly, the fitness of each individual in the population is calculated according to the current population situation. Secondly, individuals in the current population are ranked according to their fitness and diversity contribution. Lastly, parents in the mutation operators are selected in proportion to their rankings based on fitness and diversity. Thus, promising individuals with better fitness and diversity are more likely to be selected as parents. The MS-CDE framework can be applied to most CDE variants. In this study, the framework is applied to two popular representative CDE variants, (μ + λ)-CDE and ECHT-DE. Experiment results on 24 benchmark functions from CEC’2006 and 18 benchmark functions from CEC’2010 show that the proposed framework is an effective approach to enhance the performance of CDE algorithms.     

Correction to: Parallel optimization of the ray-tracing algorithm based on the HPM model

The Journal of Supercomputing - A correction to this paper has been published: https://doi.org/10.1007/s11227-021-03680-0