三省少年-CSDN博客

原创 Deep Learning for Joint Channel Estimation and Feedback in Massive MIMO Systems

这是一篇简单、有创新的文章。应该是我知道的第一篇讨论联合信道估计和信道反馈的文章，接下来就看看这篇文章讲了什么吧。

2021-06-19 23:34:38 585

原创 Deep learning-based CSI Feedback for Beamforming 2

system modelsingle-cell modely=hHx+ny=\mathbf{h}^H\mathbf{x}+ny=hHx+nwhere h∈CNt×1\mathbf{h}\in \mathbb{C}^{N_t\times 1}h∈CNt×1 is the downlink channel between the BS and the user, ???? is the complex additive Gaussian noise (AWGN) with zero mean and v

2021-01-06 11:49:19 393

原创 Deep learning-based CSI Feedback for Beamforming 1

1.AbstractThe potentials of massive multiple-input multipleoutput (MIMO) are all based on the available instantaneous channel state information (CSI) at the base station (BS). Therefore, the user in frequency-division duplexing (FDD) systems has to keep

2020-12-30 23:53:26 653

原创 Iterative Algorithm Induced Deep-Unfolding Neural NetworksPrecoding Design for MultiuserMIMOSystem 1

1.AbstractAbstract—Optimization theory assisted algorithms have received great attention for precoding design in multiuser multiple-input multiple output (MU-MIMO) systems. Although the resultant optimization algorithms are able to provide excellent perfo

2020-12-25 16:26:44 646

原创 Deep learning for joint channel estimation and feedback in massive MIMO systems(2)

3.channel estimation and feedback networkIn this section, two joint channel estimation and feedback networks will be developed, where quantization method will also be discussed. Training strategy of each network will also be described at the end of this s

2020-12-19 21:31:42 357 1

原创 Deep Learning for joint channel estimation and feedback in massive MIMO systems(1)

1.IntroductionIn this paper, a comprehensive research is carried out to establish a joint channel estimation and feedback framework for the FDD massive MIMO systems based on DL techniques. The main contributions of this paper are summarized as follows.T

2020-12-17 23:59:53 495 2

原创对于Y=Hx的H细节的一些讨论

今天看到一篇论文，system model是base station(BS)有NtN_tNt根天线，user equipment(UE)是单天线,OFDM有NcN_cNc个载波将MIMO-OFDM的信道表示为[1H23456789] \left[ \begin{matrix} _1^H& 2 & 3 \\ 4 & 5 & 6 \\ 7 & 8 & 9 \end{matrix} \right] ⎣⎡1H4725836

2020-11-27 23:57:04 2490 1

原创卷积神经网络CNN

Content1.引入2.边缘检测一些工具1.矩阵灰度图化代码1.引入思考这样一个问题：我们要对100010003的图片进行分类，怎么用机器学习完成这样的工作呢？如果用全连接网络（FCN）来实现，这意味着第一层网络会有3×1063\times 10^63×106这么多个节点，网络的参数量将会是不可接受的。基于此，我们开始研究新的网络2.边缘检测对于前面提到的图像分类问题，我们一般会检测出其边缘，据此来确定图形长什么样子。边缘检测分为垂直边缘检测和水平边缘检测，怎么做边缘检测呢？假设有小的灰度图片

2020-11-11 22:07:13 306 3

原创支持向量机(Support Vector Machine SVM)

1.摘要这篇文章主要介绍支持向量机的有关内容，主要包括支持向量机的推导，支持向量机的特性，支持向量机简单的数学原理介绍，核函数，最后还对比了一下逻辑回归，支持向量机，神经网络，简单说明他们各自的适用情况。关键词：支持向量机，核函数。2.正文2.1支持向量机的引入推导支持向量机是一种强大的拟合非线性方程的工具，首先介绍一下它的引入。由前面学习，我们可以知道，对于逻辑回归的一组样本，我们有代价函数：J(θ)=−ylog11+eθTx−(1−y)log(1−11+eθTx)J(\theta)=-ylo

2020-08-19 23:02:33 761

原创简单机器学习系统的构建以及对于不对称性的分类介绍和性能评价

1.前言这篇文章主要是总结简单机器学习系统的构建，简单的误差分析，特别的，对于不对称性分类的误差评估做了介绍，并对这种情况做了性能指标的规定。关键词：机器学习系统不对称性分类查准率召回率目录1.前言2.内容介绍2.1构建机器学习系统的简单例子2.2不对称性分类的误差评估2.2.1不对称性类的介绍2.2.2新的误差评估方法2.2.3查准率和召回率的权衡2.2.4评价指标3.reference2.内容介绍2.1构建机器学习系统的简单例子假设我们需要做一个垃圾邮件屏蔽器，要怎么开始做呢？首先

2020-08-16 21:43:02 523

原创机器学习算法性能评价，改进的理论基础以及改进措施

在开始介绍这部分内容之前，首先要给出吴恩达在课上的一段话So if you like you can to think of this as a way of avoiding whats sometimes called premature optimization in computer programming. And this idea that says we should let evidence guide our decisions on where to spend our time

2020-08-12 16:34:13 702

原创谈谈考虑了多普勒的信道

无线通信中，所有的努力都是为了解决信道带来的不确定性。但是尝尝看到一些文章的信道只是瑞利多径信道，这样的信道是最简单的。这篇文章主要谈谈复杂一些的信道——加了多普勒的信道。1.多径瑞利信道所谓多径瑞利信道，一般的数学表达式为h=∑i=1Phiδ(t−τi),h=\sum_{i=1}^{P}h_i\delta(t-\tau_i),h=i=1∑Phiδ(t−τi),其中iii代表多径，根据不同的多径时延，我们会给不同的径分配不同的功率，一般是随着时延的增加功率系数会递减，从而形成我们的多径瑞利信道。这

2020-07-31 00:11:24 3479 2

原创信道模型时延和路径增益：cost207,cost2100,cost259,cost273,SCM,I-METRA,WINNER II

1.cost207适用于远郊地区（RA），典型城区（TU），恶劣城区（BU）和丘陵地区（HT）。给出了时延功率谱和多普勒功率谱，时延功率谱表明随着时间变长，逐渐下降，对于多普勒功率谱，Gauss I或者Gauss II功率谱分布只有一个多普勒频移，但是Jakes和Rice的功率谱分布会有两个峰，也就是有两个多普勒频移。使用Gauss I或者Gauss II功率谱分布的条件是delay必须大于0.5μs\mu sμs。使用各个多普勒功率谱的条件如下1.1远郊地区（RA）：6路径信道模型参数1

2020-07-23 22:38:38 9414 1

原创用python处理txt文件

1. 做法处理这样子的txt文件，需要把误码率读出来import linecacheimport docximport reimport numpy as np#count = linecache.getline("C:\\Users\\admin\\Desktop\\470_0.003.txt",4)list=[]#a=len(open(r"C:\\Users\\admin\\Desktop\\470_0.003.txt",'rU').readlines())#a=len(open(r

2020-07-19 23:56:34 1697

原创 Machine Learning for Communication Networks

Introduction越读越觉的写的很好，写篇文章给大家分享一下Main partIn order to exemplify applications of supervised and unsupervised learning, we will offer annotated pointers to the literature on machine learning for communication systems. Rather than striving for a comprehens

2020-06-18 15:57:20 743

原创 OTFS Channel Estimation(1)

content1.方法介绍1.1 method in "MIMO-OTFS in High-Doppler Fading Channels: Signal Detection and Channel Estimation"1.2 method in "On OTFS Modulation for High-Doppler Fading Channels"1.3 method in other two papers2.一个疑问reference1.方法介绍1.1 method in “MIMO-OTFS

2020-06-05 15:09:55 4505 11

原创 Delay-Doppler equalization(8)(时延多普勒均衡)⭐

content1.introduction2.difficulty3.solution不得不说，paper写的是越来越入佳境了。1.introductionIn the uplink direction, the streams from the different users arrive to the base-station when they are superimposed on one another and the base-station must separate between t

2020-05-26 23:17:08 3372

原创 Multicarrier Interpretation of OTFS（OTFS的多载波解释）（7）⭐

content1.overview(概述)2.Details(细节)2.1Introduction of delay-doppler grid and time-frequency grid(时延多普勒网格和时间频率网格的介绍)2.2Fourier relation between the two grids(两个网格之间的傅里叶关系)2.3About symplectic exponential function(关于辛指数函数)3.Conclusion1.overview(概述)In this se

2020-05-24 23:24:34 2674

原创 The delay-doppler Channel Symbol Coupling(延迟-多普勒信道符号耦合)(6)

content1.Definition定义2.Example例子2.1Channel symbol coupling of the TDM pulses2.2Channel symbol coupling of the TDM pulses2.3Channel symbol coupling of the OTFS pulses前言：和前面的类似，这部分依旧比较偏概念性的描述而缺少实际的公式和例子说明，是对延迟-多普勒信道符号耦合特性的描述1.Definition定义The wireless chan

2020-05-22 22:53:15 2608 2

原创 The OTFS Waveform&The Delay-Doppler Channel Symbol Coupling(5)

1.premiseThe grid defined in this way consists of ???? points along the delay period, with spacing Δ???? and ???? points along the Doppler period, with spacing ????????, resulting with a total of ???????? grid points inside the fundamental rectangular dom

2020-05-21 22:52:42 451

原创 The General Framework Of Signal Processing&OTFS Modulation Scheme(信号处理的一般框架&OTFS调制)（4）

content2.3 The General Framework Of Signal Processing2.4 OTFS Modulation阅前说明：这两个部分我感觉paper的着墨不多，在说核心思想。但是既无例子，也无推导过程，尤其是OTFS Modulation这个部分更是只说了核心的部分，比起前一篇论文OTFS的调制过程显得很空。看看后面会不会详细补充吧2.3 The General Framework Of Signal ProcessingThe general framework of

2020-05-18 23:22:28 1582

原创 The Delay-Doppler Signal Representation(时移多普勒信号表示)(3)

contentChannel Representation in Delay-DopplerSignal Representation in Delay-DopplerConversion among different representationChannel Representation in Delay-DopplerIn communication, they are used to represent channels by means of a superposition of time

2020-05-16 23:21:56 2217 1

原创 OTFS: A New Generation of Modulation Addressing the Challenges of 5G(2)

content2.1 OTFS in a Nutshell2.1.1 overview2.1.2 Three fundamental properties2.2 The Delay-Doppler Signal Reprsentation2.1 OTFS in a Nutshell2.1.1 overviewOTFS is a modulation scheme that multiplexes QAM information symbols in a new signal representati

2020-05-12 23:20:07 1546

原创多小区下小区上行速率的计算(6）

发现问题： result2(x)=first/(second+third+forth+1); x=x+1; end result1=result1+result2;依然是信噪比求和，还没有改过来。之前做的下行的大概在12左右终于来到写自己方法的关键时刻了吗！！...

2020-05-12 22:38:38 303

原创多小区下小区上行速率的计算(5）

今天要写时移上行速率。 d_in=rand(1,K)*0.99+0.01; theta_in=rand(1,K)*(2*pi); d_out=rand(6,K)*0.99+0.01; theta_out=rand(6,K)*(pi);小区内和相邻小区随机用户的产生，check。 alpha_1=pr*K*z*d_in(j)^r; alpha_2=6*pf*z*2^(-3.8); alpha=alph

2020-05-11 21:43:40 371

原创 OTFS: A New Generation of Modulation Addressing the Challenges of 5G(1)

1.abstractIn this paper, we introduce a new 2D modulation scheme referred to as OTFS (Orthogonal Time Frequency & Space) that multiplexes information QAM symbols over new class of carrier waveforms that correspond to localized pulses in a signal repre

2020-05-11 15:30:08 1580

原创 MIMO-OTFS in High-Doppler Fading Channels:Signal Detection and Channel Estimation(5)

5 Channel Estimation5.1 A factfinite support是什么意思呢？观察上面这幅图，我们可以知道我们调制的时延多普勒域是[0,0]−[32,32],[0,0]-[32,32],[0,0]−[32,32],但是信道的响应只有[0,0]−[5,5][0,0]-[5,5][0,0]−[5,5],它的含义是信道响应所影响的时延多普勒域只占到我们调制的多普勒域的一小部分。5.2 example这个例子表达的意思很简单，我们在(np,mp)(n_p,m_p)(np,mp

2020-05-09 19:45:49 1357 1

原创 MIMO-OTFS in High-Doppler Fading Channels:Signal Detection and Channel Estimation(4)

5MIMO-OTFS Signal Detection5.1Algorithm for MIMO-OTFS signal detection首先使用的是最大后验概率判决准则，这种检测具有指数级的复杂性。对于symbol by symbol最大后验概率判决准则有这样的检测方法因为较高的复杂度，我们采用了基于MP(message passing)的方法。对于paper中的检测方法，初次阅...

2020-05-02 23:32:45 1666 6

原创多小区下小区上行速率的计算(4）

question1现在遇到的问题时仿真出来的速率偏低。在与前面的速率对比时，突然发现自己竟然把信噪比的和加起来求和速率。修改代码修改代码 forth=forth_1+forth_2; result2(x)=first/(second+third+forth+1); x=x+1; end ra...

2020-05-02 17:34:38 398

原创 MIMO-OTFS in High-Doppler Fading Channels:Signal Detection and Channel Estimation(3)

content3.3Vectorized formulation of the input-output relation4MIMO-OTFS Modulation3.3Vectorized formulation of the input-output relation信道的脉冲响应可以写为h(τ,ν)=∑i=1Phiδ(τ−τi)δ(ν−νi)h(\tau,\nu)=\sum_{i=1}...

2020-05-01 22:57:52 621

原创多小区下小区上行速率的计算(3）

接着为了确定计算小区速率在空分上出了什么问题，我们将这种计算小区速率的思想运用到时间同步的方法中来和paper的结果作对比，以便找出其中的问题。clear;SNR=0;result=zeros(1,10);for i=1:10000 result1=zeros(1,10); pu=10; %pr=10db pr=10; r=-3.8; K=1...

2020-05-01 21:10:07 431

原创多小区下小区上行速率的计算(2)

顺着昨天的思路，明显可以看到速率偏低，于是运行代码，发现第四项的结果非常大，仔细一看发现距离产生的有问题，修改距离的产生 d_out_ture=sqrt(2^2+d_out.^2-2*2*d_out.*cos(theta_out)); forth_2=sum(pu*z*d_out_ture(1,:).^r)-pu*z*d_out_ture(1,j)^...

2020-05-01 09:38:21 301

原创 MIMO-OTFS in High-Doppler Fading Channels:Signal Detection and Channel Estimation(2)

content3.2 OTFS modulation3.2 OTFS modulation上一节阅读到了TF modulation的部分，也就是上图中的小方框，今天将继续阅读 OTFS modulation。首先我们要做的，就是把x[k,l]x[k,l]x[k,l]做ISFFT&transmit windowing变换为X[n,m]X[n,m]X[n,m]ISFFT(inver...

2020-05-01 09:16:08 890

原创 MIMO-OTFS in High-Doppler Fading Channels:Signal Detection and Channel Estimation(1)

content1.abstract2.introduction3.OTFS Modulation3.1 Time-frequency modulation1.abstractwe focus on MIMO-OTFS which brings in the high spectral and energy efficiency benefits of MIMO and the robustne...

2020-04-29 23:04:18 1473 5

原创多小区下小区上行速率的计算的想法

问题：小区速率比paper中给出的仿真速率高很多。之前小区速率计算的逻辑，是循环一万次计算出一个用户的速率，然后乘用户数。现在想的修改思路是直接每次将小区速率求出来，然后取循环一万次后的平均值。后者我觉得在一定程度上可以减小速率的随机性。因为仿真的结果表明，仿真速率的差距就在同小区的大尺度衰落上面，随机产生的大尺度衰落差距会很大。但是如果我们考虑整个小区所有用户的速率然后加和的话，大的速率固然比...

2020-04-29 20:07:44 317

原创 How to use neural network to realize logic 'and' and 'or'?

content1.Realize logic 'and'1.1precondition1.2way to realize it1.Realize logic ‘and’1.1precondition我们有两个输入，分别是x1x_1x1&x2x_2x2,且他们只能取0或者1.神经网络的模型为同时，我们有hθ(x)=11+eθTxh_{\theta}(x)=\frac{1}{1...

2020-03-31 15:07:08 169

原创 Model Representation--machine learning

content1.How does the nervous system work2.What do we learn from nervous system?1.How does the nervous system work这是一个简单的模型，一个神经元接受来自其他神经元的输入，并且输出信息到其他神经元。考虑到神经元之间的联系，我们有更加复杂的模型几个神经元作为一个神经元的输入，这个...

2020-03-27 22:51:36 187

原创 java运行matlab代码

@toc今天老师让重新部署之前做过的一个项目，项目需要用java来调用matlab代码，之前隐约记得用小窗模式调用matlab的时候只能调用单个m文件，也就是说若m文件调用写在其他m文件中的函数的这种m文件是无法运行的。那么今天的工作将按以下流程展开确认16QAM的文件是否可以被调用若是可以调用，完善代码，添加内容；若是不能调用，则建议修改例程problem1windows批处理m...

2020-03-08 16:28:15 565

原创 Neurons and the brains

Neurons and the brains1.what is it?2.why have one learning algorithm?3.why we need deep learning?3.1example11.what is it?Origins:Algorithms that try to mimic the brainWas very widely uesd in 80s a...

2020-03-03 17:15:26 174

原创 Become Happier by Learning & Applying Psychological

1.abstractScience of happinessPractice of happinessWhat makes you happymisconceptions about happinesswhy our expectations are so badwhy are we not accurate about the kinds of things that are g...

2020-03-01 22:09:22 238