A subspace detection method of analog space-time codes for multiantenna ultra-wideband transmissions

In this letter, we propose a subspace based detection method for space-time block codes (STBC) wedded with ultra-wideband (UWB) transmissions. Without the need of channel information, the proposed algorithm yields the estimation of transmitted symbols by minimizing some quadratic form built on the orthogonality between signal and noise subspaces. Simulations in flat-faded application scenarios show that the subspace method can achieve the same diversity and a loss of about 2 dB at the 10/sup -3/ level with more than four successive space-time codes being decoded jointly, compared to the coherent decoding algorithm.     

基于注意力机制面向短文本多分类的关键词权重优化

关键词语义敏感影响短文本选择关键词赋予合适权重.针对仅关注关键词是否完备没有考虑到混淆关键词会对分类造成消极影响的问题,提出一种降低混淆关键词权重实现关键词权重优化的模型.首先,基于词频?逆文件频率(TF-IDF)和混淆矩阵的原理选择文本中被定义的混淆关键词.然后,基于注意力机制构建文本表征,通过全连接层降维重构表征;训练重构的表征尽可能相似于原表征,从而选出能保留句子信息的关键词;将混淆关键词从提取的关键词里排除出去,将筛选后的关键词称为强关键词.最后,使用双向长短记忆网络?注意力机制(BiLSTM-Attention)经典模型作短文本多分类的基础模型.在此基础模型之上把强关键词整体做嵌入表示加入到BiLSTM-Attention模型中的注意力部分激活函数计算当中.与BiLSTM-Attention基础模型进行的实验结果表明,所提模型在Snippets公开数据集上,分类准确率提高0.41个百分点.     

Efficient embedded wavelet codec based on list structure

A new efficient image codec called embedded wavelet coding based on list structure (EWCBL) is proposed in this paper. By combining the techniques of morphological representation and quadtree partitioning, EWCBL effectively exploits both within-subband clustering and cross-subband similarity of wavelet coefficients. Based on the list structure, a fine fractional bit-plane coding is employed to achieve excellent rate-distortion performance. The patterned morphological dilation is introduced to greatly reduce the redundant test operations. Experimental results show that the proposed EWCBL outperforms state-of-the-art embedded codecs for both lossy and lossless compression. Moreover, the good performance and features of embeddedness and scalability are achieved with relatively low complexity.     

构造短码长不规则LDPC码

在给定长度和度数分布对情况下,不规则短码长LDPC码的性能差别很大。本文给出了一种低复杂度的基于树的贪婪搜索算法(GTS),此算法通过调整校验矩阵中环的分布情况,尽量减少短长度环,获得最大的平均最小环,从而构造性能很好的短码长不规则LDPC码。     

基于二次规划的2D FIR滤波器设计

本文提出了一种新型的基于McClenllan变换和二次规划的2D FIR滤波器设计方法.该方法使用McClenllan变换来实现从1D原型滤波器向2D滤波器的频率映射.在设计过程中,使用二次规划方法来求解McClenllan变换的最优系数,通过对二次规划使用一些限制条件,避免了传统的尺度变换问题.与传统的2D FIR滤波器设计方法相比,该方法使映射中1D FIR通带截止频率和阻带截止频率两点误差平方和达到最小,通过实际的扇形滤波器和菱形滤波器设计证明,本文的基于二次规划的方法是一种有效的2D FIR滤波器设计方法.     

超软赝势密度泛函分子动力学计算中的若干优化算法

基于密度泛函理论的第一性原理计算是科学计算中重要的领域,被广泛用于材料计算和量子化学等研究。我们在PEtot软件的基础上研发了GPU加速软件PWMat,本文主要介绍:1)针对超软赝势密度泛函算法部分的GPU算法优化与实现,取得了2-3倍的加速;2)针对超软赝势密度泛函分子动力学计算部分的若干算法优化,包括原子受力求解算法的修正、电荷密度和波函数的插值算法的改进等。其中原子受力的收敛精度获得了1-2个数量级的提高,电荷密度的收敛精度获得了2-4个数量级的提高。     

Analysis of frequency selective surfaces through the blazing onsetusing rational Krylov model-order reduction and Woodbury singularityextraction

Rational Krylov model-order reduction techniques are used to accelerate the characterization of frequency selective surfaces (FSSs) over broad spectral ranges, including frequencies at which a higher order Floquet mode stops evanescing and begins to propagate. The procedure comprises three main stages: (1) construction of the spectral Galerkin system at a small set of frequencies, (2) linearization of that system, and (3) reduction of the linearized system using the rational Krylov technique. The inclusion of blazing frequencies in the band of interest complicates the second and third of these steps because of the branch point singularity in the periodic Green's function. This difficulty is avoided by removing the blazing modes contributions to the spectral Galerkin matrix using the Woodbury formula for low-rank updates. The algorithm results in a set of small linear systems producing outputs that are combined to approximate the reflection and transmission coefficients of all propagating modes. The technique is applied to three different frequency selective surfaces and is shown to be accurate and efficient in all cases     

Analysis of frequency selective surfaces using two-parametergeneralized rational Krylov model-order reduction

Generalized rational Krylov model-order reduction techniques are applied to the spectral Galerkin system describing frequency selective surfaces, resulting in surface reflection coefficient models that depend on both the frequency and the incident angle of the exciting wave. The procedure is composed of three steps: construction of the spectral Galerkin system, linearization of that system, and reduction of the linearized system. The linearization of the spectral Galerkin matrix is carried out using two-dimensional (2-D) polynomial interpolation and the generation of a “two-parameter” companion form of the polynomial system. The subsequent model-order reduction is based on the concept of generalized Krylov subspaces, which are defined in the text. It results in a small system with a frequency and incident angle dependent output that matches the two-parameter polynomial interpolant system transfer function and its derivatives at many points in the frequency/incident angle plane. The technique is applied to the characterization of several frequency selective surfaces, and numerical results that demonstrate the accuracy of the techniques over a broad band of frequencies and range of incident angles are presented     

Reduced-order modeling of multiscreen frequency-selective surfacesusing Krylov-based rational interpolation

A method is presented for generating a broad-band rational interpolant approximation of the reflection coefficient of multiple-screen frequency-selective surfaces (FSSs). The technique is structured around a linearization of the system provided by a spectral domain moment method-based analysis of the FSS, followed by a model-order reduction of the linearized system using the dual rational Arnoldi method. This process creates a rational interpolant of the linearized system that matches its transfer function and its derivatives at several expansion points in the Laplace domain. Numerical results indicate that a reduced-order model with a system matrix of dimension less than 20×20 can accurately reproduce the broad-band behavior of multiscreen FSSs originally modeled with several hundreds or thousands of unknowns     

Two-dimensional transverse-magnetic time-domain scattering using the Nystro/spl uml/m method and bandlimited extrapolation

A method is presented for the solution of the integral equations that describe the electromagnetic scattering from an infinitely long conducting cylinder in the time domain. The method discretizes the integral equations spatially using a high-order locally-corrected Nystro/spl uml/m method and temporally using a filtered kernel. The filtering of the kernel both controls aliasing and reduces the order of its singularity. On the other hand, filtering also gives rise to a noncausal kernel so the time marching is accomplished with a bandlimited extrapolation scheme. Numerical results demonstrate the stability and accuracy of the proposed method.