高原公路纵断面线形与驾驶员脑电特性的试验分析

为了解驾驶员在高原公路行车时,其脑电特性与纵断面线形的关系,探究适宜的公路线形,选在帕米尔高原G314国道进行驾驶试验。通过对比经过181个变坡点附近受测驾驶员脑电(EEG)的变化特征,选取脑电α频段与β频段的平均功率比R作为评价驾驶员心理紧张度的脑电特性指标,定量分析R随海拔、竖曲线半径r、坡差ω的变化规律;同时建立关系模型。研究表明:R与海拔、竖曲线半径、坡差之间的相关性均较强;R与海拔、绝对坡差呈正相关,与r呈负相关。表明海拔的升高,绝对坡差的增大,竖曲线半径的减小,对驾驶活动是不利的,驾驶员的生理不适状况会随之逐渐增加,疲劳感逐渐增强。     

UKF和UPF在惯导初始对准中的应用研究对比

捷联惯性导航系统是新型航位推算系统,在惯导系统执行工作任务之前需要进行初始对准,以保证系统的正常运行。对捷联式惯导系统,初始对准就是确定初始时刻的姿态阵,利用惯性元件的输出信息,选用合适的滤波方法,将计算的导航坐标系与真实导航坐标系的失准角估计出来,来修正姿态矩阵,使计算坐标系与真实坐标系尽可能重合。在实际的导航系统中,状态方程和量测方程通常都是非线性的,对于非线性特性,传统的解决方法是利用EKF滤波算法,但它只适用于弱非线性模型的估计,系统的非线性越强,引起的估计误差就越大,甚至会引起滤波发散。为此提出两种滤波算法UKF与UPF,并将两者进行了仿真对比,结果表明UPF算法比UKF算法收敛速度更快,估计精度更高。     

刚构-连续组合体系梁桥边跨合龙方案

以东路高架桥边跨合龙段施工工程为背景,采用平面杆系程序对边跨合龙方案由落地支架施工更改为吊架施工的桥梁结构进行了对比分析.研究结果表明,采用不同的边跨合龙方案对主梁在施工阶段、成桥阶段的受力状态影响较小.并对当过渡墩较高或支架需搭设在水中时,采用吊架施工的可行性进行了探讨.     

射影几何观点下二阶曲线度量性质的研究

为了在高观念下审视初等几何提供理论支撑,运用射影几何的方法研究了二阶曲线的主轴、顶点、焦点、准线、离心率等度量性质.得到三种类型的非退化二阶曲线其主轴、顶点、焦点的个数及离心率的大小等情况.     

语音信号时间动态规正新方法

利用函数逼近原理，提出了一种语音信号时间动态规正的新方法，并由此导出了可用于神经网络语音识别的新特征。重点讨论了时间动态规正的原理、新特征的提取方法和性质。实验及理论分析表明，新特征对于神经网络语音识别及语音压缩具有很好的效果。     

SERVICE VALUE NETWORKS： INTO PRACTICE

Modeling a business to match its customer perceived （and customer desired） outcomes, remains an on-going task. This paper considers shows how a service value networks （SVNs） approach may be engaged to model, and deliver, understanding of the front-end business and its direct engagement with its immediate （or front-end） customers. These front-end customers may be either： （1） its off-line, in-store customers - engaging directly with the business sales staff （or representative）, or-（2） its on-line, virtual customers - engaging with the business via internet or remote access. In this situation, both the business, and the customer draw upon their respective viewpoints, and both sides influence the interaction. External to these business and customer influences are additional factors that capture the immediate and broader global effects - termed environmental effects. These three business-customer engagement areas of influence are captured by a SVN SEM approach.
Without a business-customer encounter of some kind the likelihood of a services business, and a prospecting customer, successfully engaging in an exchange process is reduced. This exchange may be a physical, and/or services exchange, and/or an information and/or ideas exchange. To the business, this encounter is, in effect, a trade, and as a result, the business targets acquiring an economic exchange that will ultimately deliver a net positive economic outcome. To the customer, external and internal information feeds, servicing, perceived value, and satisfaction, are key encounter drivers. SVNs offers a new way to understand the business-customer encounter, and to then utilize this acquired knowledge to either： （1） build a strategic management model, or （2） reengineer its business networks, and to then create a re-positioned, more customer-aligned business operation - one that is capable of delivering win - win, competitive business-customer solutions.     

A Research on BIlingual Dictionary Based Sentence Alignment for Chinese English Parallel Corpus

Parallel corpus is of great importance to machine translation,and automatic sentence alignment is the first step towards its processing.This paper puts forward a bilingual dictionary based sentence alignment method for Chinese English parallel corpus,which differs from previous length based algorithm in its knowledge-rich approach.Experimental result shows that this method produces over 93% accuracy with usual English-Chinese dictonaries whose translations cover 31.88%-47.90% of the corpus.     

基于VPU加速的嵌入式实时人脸检测系统设计与实现

智能设备高昂的设计费用和庞大的计算资源需求成为在便携式、低功耗设备上实现深度学习算法及其应用的主要障碍.文中基于树莓派平台,借助Intel的视频处理器(V PU)低功耗加速模块,设计并实现了基于残差特征提取模块CNN模型的实时人脸检测系统.结果表明,相较于单纯使用树莓派CPU进行计算,文中方法在视频流中检测人脸和人脸关键点提取的实验中分别实现了18.62倍和17.46倍的加速,在便携式设备中实现快速、实时、在线的人脸检测和特征点提取成为现实,同时为使用便携式、低功耗设备运行深度学习算法提供了一种确实可行的方案.     

多注意力结合光流的视频超分辨方法

针对现如今监控摄像、卫星遥感以及视频娱乐等领域对视频图像的清晰度要求越来越高,而目 前大部分视频超分辨方法存在参数量大、恢复的视频存在抖动等问题,提出了一种多注意力结合光流的视频 超分辨方法,通过引入多个注意力包括空间注意力、通道注意力以及自注意力来提升超分辨性能。 具体而 言,作为一种特征加权的增强方法,这些注意力可以捕获视频帧的时空特征并增强自适应性和通道间的依赖 性,实现全局学习的功能;同时,提出双阶段特征对齐思路,首先利用光流对视频进行估计,进行第一阶段的 特征对齐,然后引入长短是记忆网络结构增强位置和通道的特征融合,进行第二阶段的特征对齐,以防止恢 复的视频帧出现抖动。 实验结果表明:该方法在评估标准和可视化效果方面都取得了令人满意的效果。     

Grinding Characteristics Of Directionally Aligned SiC Whisker Wheel-Comparison With Al2O3 Fiber Wheel

A unique SiC whisker wheel was invented,in which the whiskers were aligned normally to the grinding wheel surface.In this paper,grindabilities of the SiC whisker wheel are investigated and compared with those of other wheels of SiC grains,Al2O3 grains,as well as Al2O3 long and short fibres which were also aligned normally to the grinding wheel surface,respectively.The main research contents concern grinding characteristics of a directionally aligned SiC whisker wheel such as material-removal volume,wheel-wear rates,integrity of the ground surfaces,grinding ratios and grinding efficiency.Furthermore,grinding wheels of whiskers and fibres have a common disadvantage:they tend to load easily.The authors have proposed a simple method of loading-free grinding to overcome this propensity and investigate some related grinding characteristics under loading-free grinding conditions.