In recent years, arbitrary style transfer (AST) with high quality develops rapidly and the application is widespread. Customers desire to stylize their portraits and edit face attributes simultaneously. To satisfy people's demands, we design a whole framework to complete both global style transfer in color and local style transfer in attributes. First, a target attribute enhancement module guided by the latent space factorization is presented to alleviate over-editing problems and obtain style codes. Then, a local-global fusion module is proposed to integrate target style codes and global features. Besides, a contrastive coherence preserving loss is built in a pre-trained style transfer network to better adapt to face images. The experiments show that our model can generate pleasant images.

Although generative models have made great progress in art image generation, few of them pay attention to sketch-based art image generation. Generating art images from sketch is a challenging problem suffering from two main issues: (1) how to constrain the generation of art images with sparse sketch features and (2) how to fully utilize the style information of a reference image without being influenced by their content features. To tackle these, we propose a GAN-based method for art image generation given the sketch and reference style. Specifically, a style feature enhancement module and a sketch-adaptive normalization module are constructed to enable the disentanglement of the content information from the reference style image. Experiments and comparisons demonstrate the superiority of our model over current generative models in sketch-based art image generation.

Patients with chronic kidney disease (CKD) may undergo cognitive impairment. We aimed to explore the cognition of patients with cognitive impairment (CI) and no cognitive impairment (NCI) respectively and the effect of demographics, estimated glomerular filtration rate (eGFR), number of comorbidities (NCD), and hemoglobin on CI in Chinese patients with CKD at stage 3-5 treated by nondialysis by using the Beijing version of the Montreal Cognitive Assessment (MoCA-BJ). A total of 120 patients with CKD were recruited from the Department of Nephrology at the Affiliated Hospital of Nanjing University of Chinese Medicine at in-patient and out-patient follow up. A logistic regression model was performed to assess the effect of these variables on CI of CKD patients. The results indicated that the CI group was mainly in the decline of visuospatial and executive function, abstraction, and memory, compared with the NCI group. In addition, years of education, eGFR and NCD were found as predictors of CI of CKD patients at stage 3-5. Specifically, lower eGFR, less years of education and more comorbidities were risk predictors of CI.

Irreversible electroporation (IRE) is a novel tumor ablation technology that applies an external pulsed electric field to generate nanoscale irreversible pores in the tumor cell membranes, thereby destroying cellular homeostasis and inducing apoptosis. It has some unique advantages over thermal ablations, including no sensitive to the `heat-sink' effect of blood vessels and preserving surrounding vital structures. However, there is still a risk of cancer recurrence with IRE, especially with heterogeneous tissues. For heterogeneous tissues, the existence of structures with different conductivities makes it difficult to control the ablation outcomes of IRE. In this study, we used finite element simulation combined with experiments to observe the influence of heterogeneous implants on the shape of the ablation zone in a plant model. Electrochemical impedance spectroscopy was used to extract the impedance data and characterize the relationship between relevant electrical parameters and ablation area. We found that the pulsed electric field would be distorted by the heterogeneous implant with high electrical conductivity, resulting in the irregular shape of the ablation zone, and the relative change value of the extracellular fluid equivalent resistance extracted from impedance spectrum was linearly related to the ablation zone. In conclusion, the ablation zone could be predicted by the proposed method in the study with an acceptable accuracy.

Music Emotion Recognition (MER) has attracted much attention in the last decades. Many novel methods and new audio features have been designed to improve the performance of MER algorithms. However, it is difficult to apply the previous methods in cross-modal interaction tasks and emotion-based music generation tasks. To overcome this issue, we conduct experiments on the 4Q audio emotion dataset for better application of MER. Furthermore, we propose a novel Attention-based Joint Feature Extraction Model (AJFEM) for high performance predictions in static MER. In the feature extraction module, we utilize the filter bank and log mel spectrograms to extract features. Specially, attention mechanism has been applied to extract emotion-related features. Then the automatically extracted features are input into GRU-SVM to get the classification results. The experimental results show that our model is superior to most of the compared methods, and has the potential to be applied to emotional interaction tasks.

With the trend of people expressing opinions and emotions via images online, increasing attention has been paid to affective analysis of visual content. Traditional image affective analysis mainly focuses on single-label classification, but an image usually evokes multiple emotions. To this end, emotion distribution learning is proposed to describe emotions more explicitly. However, most current studies ignore the ambiguity included in emotions and the elusive correlations with complex visual features. Considering that emotions evoked by images are delivered through various visual features, and each feature in the image may have multiple emotion attributes, this paper develops a novel model that extracts multiple features and proposes an enhanced fuzzy k-nearest neighbor (EFKNN) to calculate the fuzzy emotional memberships. Specifically, the multiple visual features are converted into fuzzy emotional memberships of each feature belonging to emotion classes, which can be regarded as an intermediate representation to bridge the affective gap. Then, the fuzzy emotional memberships are fed into a fully connected neural network to learn the relationships between the fuzzy memberships and image emotion distributions. To obtain the fuzzy memberships of test images, a novel sparse learning method is introduced by learning the combination coefficients of test images and training images. Extensive experimental results on several datasets verify the superiority of our proposed approach for emotion distribution learning of images.

Although significant recent advances in condition generative model have shown remarkable improvements for controlled image generation, the image generation for multiple complex objects is still a challenge. To address the challenge, we propose a module of text description parsed into scene graph, which can generate reasonable scene layout to ensure the generated image and object realistic. Our proposed method enhances the interaction between objects and global semantics by concatenates each object embedding with text embedding To preserve the local image semantics, the Spatially-adaptive normalization(SPADE) layer is added into the generator of our model. We validate our method on Visual Genome and COCO-Stuff, where qualitative results and ablation study demonstrate the ability of our model in generating images with multiple objects and complex relationships.

In this paper, we propose an Attribute-Rich Generative Adversarial Network (AttRiGAN) for text-to-image synthesis, which enriches the simple text description by associating knowledge graph and embedding it in the synthesis task in the form of an attribute matrix. Higher fine-grained images can be synthesized with AttRiGAN, and the synthesized sample are more similar to the objects that exist in the real world, since they are driven by attributes which are enriched from the knowledge graph. The experiments conducted on two widely-used fine-grained image datasets show that our AttRiGAN allows a significant improvement in fine-grained text-to-image synthesis.

随着电磁对抗和雷达技术的不断演进,雷达信号由传统的连续波、单脉冲形式逐步向宽带线性调频、捷变频、跳频等复杂波形发展,常用的频率测量方法在测频精度和测频速度等方面很难满足要求.针对宽带相控阵雷达目标回波模拟器瞬时信号带宽高达2 GHz、扫频或随机跳频信号带宽覆盖整个工作频段的特点,创新性地采用瞬时测频引导结合实时宽带数字信道化精测频技术,设计研制了超宽带、高精度的瞬时测频模块和相应软件,并应用于宽带目标回波模拟器的研制之中.通过实测和半实物仿真试验验证,测频精度、测频范围和测频的实时性等指标完全满足整体性能要求.

In this paper, we exploit the idea of virtual ESPRIT (VESPA) to develop a multi-baseline VESPA (MB-VESPA) approach for direction finding. Specially, we define several cumulant matrices to provide ambiguous direction estimates under different baselines. Fine and unambiguous estimation is then obtained by a simple refinement step. Two refine approaches, termed as successive baseline approach and coprime baseline approach, are subsequently introduced. MB-VESPA shares all the advantages of the VESPA. It is simple, closed-form, search-free, and is applicable to irregularly linear array. In addition, it is free of the impact on the sensor gain uncertainties.

We propose a new approach which is a three-stage pipeline to fast and accurate segment hand from a single depth image. Firstly, a depth frame is segmented into several regions by histogram-based thresholds selection algorithm and tracing the exterior boundaries of objects. We found that MINIMUM, MEAN and MEDIAN are effective ways to separate objects and the threshold in the valley between two maxima similar to MINIMUM algorithm with a minimum error. Then, each segmentation proposal is evaluated by a 3-layers shallow convolutional neural network (CNN) which is trained as a binary classification function to predict whether it is a partition of hand. Finally, all hand components are merged as our hand segmentation result. In our experiment, we use a set of real data containing more than 200,000 frames of depth images. Compared with the results achieved by approaches based on RDF and SegNet, results demonstrate that our approach achieves better performance in high-accuracy (88.34% mean IoU) within shorter processing time (8 ms).

2019年12月,武汉市暴发急性传染性疾病,后经证实为新型冠状病毒( 2019-nCoV )引起的肺炎,命名为新型冠状病毒肺炎( COVID-19).该病传染性强,老年人和有慢性基础疾病者预后较差[1] ,亟待寻找有效的防治措施.我们收治了1例经恢复期血浆治疗后康复的超高龄新型冠状病毒肺炎病人,报道如下.

In this paper, an efficient self-docking and self-undocking approach that can be applied to any self-changeable robot is reported. The changeable robot is a concept to extend the concept of reconfiguration in literature, and the name is first coined in this paper. The new feature of the approach is the self-search of a target module with which the remaining robot docks with and then to complete the docking. The new feature is realized by developing modules on which there are built-in cameras and distance sensors. A deep learning method based on CNN (Convolutional Neural Network) was employed to enhance the intelligence level of the docking and undocking system. The experiment was carried out to test the effectiveness of the approach based on a proprietary self-changeable robot.

基于深度图像的手势估计比人体姿势估计更加困难,部分原因在于算法不能很好地识别同一个手势经旋转后的不同外观样式.提出了一种基于卷积神经网络(Convolutional Neural Network,CNN)推测预旋转角度的手势姿态估计改进方法:先利用自动算法标注的最佳旋转角度来训练CNN;在手势识别之前,用训练好的CNN模型回归计算出应预旋转的角度,然后再对手部深度图像进行旋转;最后采用随机决策森林(Random Decision Forest,RDF)方法对手部像素进行分类,聚类产生出手部关节位置.实验证明该方法可以减少预测的手部关节位置与准确位置之间的误差,手势姿态估计的正确率平均上升了约4.69％.

目的 探讨皮肤恶性黑色素瘤miRNA的差异表达有望为该肿瘤的诊断和靶向治疗提供依据,筛选皮肤恶性黑色素瘤血清和毛发2个部位的7个特异性miRNA表达差异.方法 收集2016年3月～2017年10月在海军军医大学附属长征医院手术局部切除并经常规组织病理学和免疫组织化学检测证实为恶性皮肤黑色素瘤(CM)的12个病例作为CM组,同期收集非肿瘤疾病的17个皮肤病病例作为对照组.采用qRT-PCR技术检测两组患者的血清和毛发中miRNA表达情况.将检测结果一致的候选miRNA确定为有意义的共同差异表达生物学指标,利用组间差异倍数筛选出差异＞2.5倍差异表达的miRNA.结果 CM组患者血清中7个miRNA有2个上调,5个下调,上调的miRNA包括miR-4487和miR-4706,下调的miRNA包括miR-16、miR-211、miR-4731、miR-509-3p和miR-514a;毛发中有3个上调,4个下调,上调的miRNA包括miR-211、miR-4487和miR-4731;下调的miRNA包括miR-16、miR-4706、miR-509-3p和miR-514a.与对照组比较,CM组miR-4487在血清和毛发中共同上调,但差异无统计学意义(P＞0.05).与对照组比较,CM组血清和毛发中has-miR-16和has-miR-509-3p表达显著下调,差异均有统计学意义(P＜0.05),并且一致性较高;与对照组比较,CM组血清和毛发中miR-541a表达虽然显示共同下调,但差异无统计学意义(P＞0.05).血清和毛发差异表达2.5倍以上的miRNA为miR-16和miR-514a;单独毛发差异表达2.5倍以上的miRNA为miR-4706;单独血清标本4倍以上差异表达为miR-4731;血清标本10倍以上差异表达的miRNA为miR-211和miR-509-3p.结论 与非肿瘤疾病患者比较,血清和毛发中存在多种miRNA的差异表达,CM患者血清中miR-211和miR-509-3p存在高度显著差异表达,可以作为该肿瘤的生物学标志物;毛发标本miR-4706的差异表达有可能成为黑色素瘤早期非创伤性检测的生物学指标.miRNA检测对CM是可利用的生物学指标,也可能成为今后靶基因治疗的手段.

Real-time hand gesture recognition technology significantly improves the user’s experience for virtual reality/augmented reality（VR/AR） applications, which relies on the identification of the orientation of the hand in captured images or videos. A new three-stage pipeline approach for fast and accurate hand segmentation for the hand from a single depth image is proposed. Firstly, a depth frame is segmented into several regions by histogrambased threshold selection algorithm and by tracing the exterior boundaries of objects after thresholding. Secondly, each segmentation proposal is evaluated by a three-layers shallow convolutional neural network（CNN） to determine whether or not the boundary is associated with the hand. Finally, all hand components are merged as the hand segmentation result. Compared with algorithms based on random decision forest（RDF）, the experimental results demonstrate that the approach achieves better performance with high-accuracy（88.34% mean intersection over union, mIoU） and a shorter processing time（≤8 ms）.

A resilient robot can recover its original function after a partial damage. In this paper, a novel under-actuated resilient robot named ReBot is presented. ReBot is an under-actuated robotic system and it consists of one type of module, which realizes the under-actuated characteristic with the whole robot. In ReBot, the module is designed as a cube structure with two rotational DOFs. The module is designed such that one can connect to another with the function of self-locking and switching between an active and a passive connection. The evolution of the configuration of an entire robotic system is performed autonomously. A preliminary experiment is carried out to show that the design requirement can be achieved.

This paper proposes an automatic editing system named Star Cut based on face recognition using deep learning and a video shot detection technique. The purpose is to establish a system for management, retrieval, and automatic recut of film and TV shots. First, the system with over 350 faces of pop stars and actors using a U-face model is trained to learn facial features. The system uses the change rate of edges to detect shot edge. After shot segmentation, the system uses the pre-trained face models to identify faces in the input film or TV drama shot by shot. Users can either choose to recognize all figures in these shots or just choose selected one to recut all the shots containing him/her together automatically. The recall rate of shot segmentation is above 95%, and the recognition rate of faces in an shooting angle of 45? is 92.45%. Test results show that the proposed system has good robustness.

In this paper we present an innovative space-variance approach named as “Space-Invariant Signature Algorithm (SISA)” for processing images from active systems, such as the liquid vibrating active system, cancer cells, tumor growth, and dead cells, for the detection and localization of abnormalities at an incipient stage. In this paper, a SISA processing algorithm is developed, and this algorithm is tested on a liquid vibrating active system. The abnormality in an active system can be defined as the obstacle or a failure, which impedes the activities such as vibrations, smooth flow of blood or electrical signals etc. Due to this impeding nature of the abnormality, some parameter perturbations are induced. In this paper using the SISA approach, these perturbations were detected in a preliminary experiment on a liquid vibrating active system. The degree and position of the space-variance helps us in the detection and localization of abnormality even at an early (incipient) stage. The space-variance signature pattern is named as `SISA signature pattern'. In the absence of any abnormality, the signature pattern is spaceinvariant, whereas, in the presence of any abnormality, the SISA signature pattern varies in the space (space-variant). The basic experimental studies on a liquid vibrating active system strongly suggest a possible use of the SISA approach as a non-invasive method for the detection and localization of abnormalities in biological tissues such as cancer cells non-invasively, and this work will be reported in another paper.

Objective To determine the effect of aspirin on the spatial learning and memory abilities and the hippocampal expression of interleukin-1β (IL-1β), tumor necrosis factor-α (TNF-α), interleukin-4 (IL-4) and interleukin-10 (IL-10) in the rat model of Alzheimer's disease (AD), and to elucidate its potential anti-inflammatory effect in the prevention and treatment of AD.Methods A total of 40 SD rats were randomly divided into 4 groups: control group, AD model group, and low-and high-dose aspirin groups (n=10).Distilled water was used to feed the rats from the 2 former groups, and 1 mg/ml and 2 mg/ml aspirin were given to the 2 latter groups respectively for 3 weeks.Then the rat model of AD was established by injecting Aβ25-35 into the lateral cerebral ventricle, and then all were fed continuously for another 3 weeks.The spatial learning and memory abilities of the rats was tested by Morris water maze.Then the rats were sacrificed and the hippocampal tissues were collected to detect the levels of IL-1β, TNF-α, IL-4 and IL-10 by enzyme linked immunosorbent assay (ELISA).Results The mean escape latency was significantly shorter in the high-dose aspirin group at the 1st, 2nd and 3rd days and in the low-dose group at the 3rd day when compared with the model group (P<0.05).The expression levels of IL-1β and TNF-α were significantly increased (P<0.001), but those of IL-4 and IL-10 were obviously decreased (P<0.05) in the model group than the control group.The low-dose group also had notably higher TNF-α and lower IL-4 levels when compared with the control group (both P<0.01).High-dose aspirin treatment resulted in obvious decreases of the IL-1β and TNF-α levels (P<0.01) and increases of those of IL-4 and IL-10 (P<0.05) than the AD group.However, low dose of aspirin only decreased IL-1β level (P<0.01).Conclusion Aspirin intervention promotes the spatial learning and memory abilities in AD rats.It exerts protective effect in the pathogenesis of AD and inhibits the development of inflammatory response through inhibiting glial activation and improving the balance of pro-and anti-inflammatory cytokines.