跳转至

第 3 章 线性代数

本章预计学习时间: 6-8 小时
前置知识: 第 1 章(概率论基础)、高中数学
后续依赖: 第 9 章(信道估计)、第 10 章(均衡算法)、第 11 章(MIMO 系统)


📌 本章目标

学完本章后,你将能够:

  1. 掌握 矩阵运算的基本规则和性质
  2. 理解 特征值分解和奇异值分解的物理意义
  3. 应用 矩阵求逆引理简化通信系统计算
  4. 推导 MIMO 系统中的关键公式
  5. 运用 线性代数工具解决信号处理问题

3.1 矩阵基础回顾

🎯 基础概念

矩阵的定义

矩阵是一个由数字排列成的矩形阵列,通常用大写字母表示:

\[ \mathbf{A} = \begin{bmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \end{bmatrix} \]
  • 维度\(m \times n\) 表示 \(m\)\(n\)
  • 元素\(a_{ij}\) 表示第 \(i\) 行第 \(j\) 列的元素
  • 方阵:当 \(m = n\) 时,称为 \(n\) 阶方阵

特殊矩阵

矩阵类型 定义 示例
单位矩阵 \(\mathbf{I}\) 对角线为 1,其余为 0 \(\begin{bmatrix}1&0\\0&1\end{bmatrix}\)
对角矩阵 \(\mathbf{D}\) 只有对角线元素非零 \(\begin{bmatrix}d_1&0\\0&d_2\end{bmatrix}\)
对称矩阵 \(\mathbf{A}^T = \mathbf{A}\) 转置等于自身 协方差矩阵
正交矩阵 \(\mathbf{Q}^T\mathbf{Q} = \mathbf{I}\) 列向量正交归一 旋转矩阵
酉矩阵 \(\mathbf{U}^H\mathbf{U} = \mathbf{I}\) 复数域的正交矩阵 DFT 矩阵
Hermitian 矩阵 \(\mathbf{A}^H = \mathbf{A}\) 共轭转置等于自身 信道相关矩阵

🔧 矩阵运算

1. 矩阵加法

$$ (\mathbf{A} + \mathbf{B}){ij} = a $$ } + b_{ij性质: 交换律、结合律

2. 矩阵乘法

\[ (\mathbf{AB})_{ij} = \sum_{k=1}^{n} a_{ik}b_{kj} \]

重要性质: - ❌ 不满足交换律\(\mathbf{AB} \neq \mathbf{BA}\)(一般情况) - ✅ 满足结合律\((\mathbf{AB})\mathbf{C} = \mathbf{A}(\mathbf{BC})\) - ✅ 满足分配律\(\mathbf{A}(\mathbf{B}+\mathbf{C}) = \mathbf{AB} + \mathbf{AC}\)

3. 矩阵转置

\[ (\mathbf{A}^T)_{ij} = a_{ji} \]

性质: - \((\mathbf{A}^T)^T = \mathbf{A}\) - \((\mathbf{AB})^T = \mathbf{B}^T\mathbf{A}^T\) - \((\mathbf{A} + \mathbf{B})^T = \mathbf{A}^T + \mathbf{B}^T\)

4. 共轭转置(Hermitian 转置)

\[ (\mathbf{A}^H)_{ij} = \overline{a_{ji}} \]

在通信中的应用: 复数信号处理、MIMO 信道矩阵

5. 矩阵求逆

\(\mathbf{A}\)\(n \times n\) 方阵且可逆,则: $$ \mathbf{A}\mathbf{A}^{-1} = \mathbf{A}^{-1}\mathbf{A} = \mathbf{I} $$

可逆条件: \(\det(\mathbf{A}) \neq 0\)

性质: - \((\mathbf{A}^{-1})^{-1} = \mathbf{A}\) - \((\mathbf{AB})^{-1} = \mathbf{B}^{-1}\mathbf{A}^{-1}\) - \((\mathbf{A}^T)^{-1} = (\mathbf{A}^{-1})^T\)


3.2 行列式与迹

📐 行列式(Determinant)

定义

对于 \(2 \times 2\) 矩阵: $$ \det(\mathbf{A}) = \begin{vmatrix} a & b \ c & d \end{vmatrix} = ad - bc $$

对于 \(3 \times 3\) 矩阵(按第一行展开): $$ \det(\mathbf{A}) = a_{11}\begin{vmatrix}a_{22}&a_{23}\a_{32}&a_{33}\end{vmatrix} - a_{12}\begin{vmatrix}a_{21}&a_{23}\a_{31}&a_{33}\end{vmatrix} + a_{13}\begin{vmatrix}a_{21}&a_{22}\a_{31}&a_{32}\end{vmatrix} $$

重要性质

  1. \(\det(\mathbf{I}) = 1\)
  2. \(\det(\mathbf{AB}) = \det(\mathbf{A})\det(\mathbf{B})\)
  3. \(\det(\mathbf{A}^T) = \det(\mathbf{A})\)
  4. \(\det(\mathbf{A}^{-1}) = \frac{1}{\det(\mathbf{A})}\)
  5. \(\det(c\mathbf{A}) = c^n \det(\mathbf{A})\)\(n\) 为矩阵维度)

物理意义

  • 几何解释:行列式的绝对值表示线性变换的"缩放因子"
  • 在通信中:MIMO 信道容量计算、矩阵可逆性判断

📊 迹(Trace)

定义

矩阵的迹是对角线元素之和: $$ \text{tr}(\mathbf{A}) = \sum_{i=1}^{n} a_{ii} $$

重要性质

  1. \(\text{tr}(\mathbf{A} + \mathbf{B}) = \text{tr}(\mathbf{A}) + \text{tr}(\mathbf{B})\)
  2. \(\text{tr}(c\mathbf{A}) = c \cdot \text{tr}(\mathbf{A})\)
  3. 循环性质\(\text{tr}(\mathbf{AB}) = \text{tr}(\mathbf{BA})\)
  4. 推广\(\text{tr}(\mathbf{ABC}) = \text{tr}(\mathbf{BCA}) = \text{tr}(\mathbf{CAB})\)
  5. \(\text{tr}(\mathbf{A}^T) = \text{tr}(\mathbf{A})\)
  6. \(\text{tr}(\mathbf{A}^H\mathbf{A}) = \sum_{i,j} |a_{ij}|^2\)(Frobenius 范数的平方)

在通信中的应用

  • 信号功率计算\(\text{tr}(\mathbf{R}_x)\) 表示信号总功率
  • MIMO 系统:信道矩阵的迹用于功率归一化
  • 估计理论:Cramér-Rao 下界计算

3.3 特征值与特征向量

🎯 核心概念

定义

对于 \(n \times n\) 方阵 \(\mathbf{A}\),如果存在非零向量 \(\mathbf{v}\) 和标量 \(\lambda\),使得: $$ \mathbf{A}\mathbf{v} = \lambda\mathbf{v} $$ 则称: - \(\lambda\) 为**特征值** - \(\mathbf{v}\) 为对应的**特征向量**

特征方程

\[ \det(\mathbf{A} - \lambda\mathbf{I}) = 0 \]

求解此方程得到 \(n\) 个特征值 \(\lambda_1, \lambda_2, \ldots, \lambda_n\)

🔍 求解步骤

例:\(\mathbf{A} = \begin{bmatrix} 4 & 1 \\ 2 & 3 \end{bmatrix}\) 的特征值和特征向量

Step 1: 写特征方程 $$ \det\begin{pmatrix} 4-\lambda & 1 \ 2 & 3-\lambda \end{pmatrix} = (4-\lambda)(3-\lambda) - 2 = \lambda^2 - 7\lambda + 10 = 0 $$

Step 2: 解特征值 $$ \lambda_1 = 5, \quad \lambda_2 = 2 $$

Step 3: 求特征向量

\(\lambda_1 = 5\): $$ (\mathbf{A} - 5\mathbf{I})\mathbf{v} = \begin{bmatrix} -1 & 1 \ 2 & -2 \end{bmatrix}\begin{bmatrix} v_1 \ v_2 \end{bmatrix} = \mathbf{0} \Rightarrow \mathbf{v}_1 = \begin{bmatrix} 1 \ 1 \end{bmatrix} $$

\(\lambda_2 = 2\): $$ (\mathbf{A} - 2\mathbf{I})\mathbf{v} = \begin{bmatrix} 2 & 1 \ 2 & 1 \end{bmatrix}\begin{bmatrix} v_1 \ v_2 \end{bmatrix} = \mathbf{0} \Rightarrow \mathbf{v}_2 = \begin{bmatrix} 1 \ -2 \end{bmatrix} $$

📐 特征值分解(Eigendecomposition)

如果 \(n \times n\) 矩阵 \(\mathbf{A}\)\(n\) 个线性无关的特征向量,则可以对角化: $$ \mathbf{A} = \mathbf{V}\mathbf{\Lambda}\mathbf{V}^{-1} $$

其中: - \(\mathbf{V} = [\mathbf{v}_1, \mathbf{v}_2, \ldots, \mathbf{v}_n]\) 是特征向量矩阵 - \(\mathbf{\Lambda} = \text{diag}(\lambda_1, \lambda_2, \ldots, \lambda_n)\) 是特征值对角矩阵

🔑 重要性质

  1. 迹与特征值\(\text{tr}(\mathbf{A}) = \sum_{i=1}^{n} \lambda_i\)
  2. 行列式与特征值\(\det(\mathbf{A}) = \prod_{i=1}^{n} \lambda_i\)
  3. 对称矩阵:特征值都是实数,特征向量正交
  4. 正定矩阵:所有特征值 \(> 0\)
  5. 半正定矩阵:所有特征值 \(\geq 0\)

📡 在通信中的应用

1. MIMO 信道容量

对于 MIMO 系统 \(\mathbf{y} = \mathbf{H}\mathbf{x} + \mathbf{n}\),信道容量: $$ C = \sum_{i=1}^{r} \log_2\left(1 + \frac{P_i \lambda_i}{\sigma^2}\right) $$ 其中 \(\lambda_i\)\(\mathbf{H}\mathbf{H}^H\) 的特征值(信道增益)

2. 主成分分析(PCA)

  • 协方差矩阵的特征值表示各主成分的方差
  • 最大特征值对应的特征向量是第一主成分方向
  • 用于降维、信道压缩

3. 功率分配(注水算法)

最优功率分配: $$ P_i = \left(\mu - \frac{\sigma2}{\lambda_i}\right)+ $$ 其中 \(\lambda_i\) 是信道特征值


3.4 奇异值分解(SVD)⭐

🎯 核心定理

任意 \(m \times n\) 矩阵 \(\mathbf{A}\) 都可以分解为: $$ \mathbf{A} = \mathbf{U}\mathbf{\Sigma}\mathbf{V}^H $$

其中: - \(\mathbf{U}\)\(m \times m\) 酉矩阵(左奇异向量) - \(\mathbf{\Sigma}\)\(m \times n\) 对角矩阵(奇异值) - \(\mathbf{V}\)\(n \times n\) 酉矩阵(右奇异向量)

📐 奇异值矩阵

\[ \mathbf{\Sigma} = \begin{bmatrix} \sigma_1 & 0 & \cdots & 0 & \cdots \\ 0 & \sigma_2 & \cdots & 0 & \cdots \\ \vdots & \vdots & \ddots & \vdots & \vdots \\ 0 & 0 & \cdots & \sigma_r & \cdots \\ \vdots & \vdots & \vdots & \vdots & \ddots \end{bmatrix} \]

其中: - \(\sigma_1 \geq \sigma_2 \geq \cdots \geq \sigma_r > 0\) 是**奇异值** - \(r = \text{rank}(\mathbf{A})\) 是矩阵的秩 - 奇异值是 \(\mathbf{A}^H\mathbf{A}\) 特征值的平方根:\(\sigma_i = \sqrt{\lambda_i}\)

🔍 SVD 的计算步骤

例:\(\mathbf{A} = \begin{bmatrix} 3 & 0 \\ 0 & -2 \end{bmatrix}\) 的 SVD

Step 1: 计算 \(\mathbf{A}^H\mathbf{A}\) $$ \mathbf{A}^H\mathbf{A} = \begin{bmatrix} 3 & 0 \ 0 & -2 \end{bmatrix}\begin{bmatrix} 3 & 0 \ 0 & -2 \end{bmatrix} = \begin{bmatrix} 9 & 0 \ 0 & 4 \end{bmatrix} $$

Step 2: 求特征值和特征向量 - \(\lambda_1 = 9 \Rightarrow \mathbf{v}_1 = [1, 0]^T\) - \(\lambda_2 = 4 \Rightarrow \mathbf{v}_2 = [0, 1]^T\)

Step 3: 计算奇异值 - \(\sigma_1 = \sqrt{9} = 3\) - \(\sigma_2 = \sqrt{4} = 2\)

Step 4: 计算左奇异向量 $$ \mathbf{u}_i = \frac{1}{\sigma_i}\mathbf{A}\mathbf{v}_i $$ - \(\mathbf{u}_1 = \frac{1}{3}\begin{bmatrix} 3 & 0 \\ 0 & -2 \end{bmatrix}\begin{bmatrix} 1 \\ 0 \end{bmatrix} = \begin{bmatrix} 1 \\ 0 \end{bmatrix}\) - \(\mathbf{u}_2 = \frac{1}{2}\begin{bmatrix} 3 & 0 \\ 0 & -2 \end{bmatrix}\begin{bmatrix} 0 \\ 1 \end{bmatrix} = \begin{bmatrix} 0 \\ -1 \end{bmatrix}\)

Step 5: 写出 SVD $$ \mathbf{A} = \begin{bmatrix} 1 & 0 \ 0 & -1 \end{bmatrix}\begin{bmatrix} 3 & 0 \ 0 & 2 \end{bmatrix}\begin{bmatrix} 1 & 0 \ 0 & 1 \end{bmatrix}^H $$

🔑 SVD 的重要性质

  1. 存在性:任意矩阵都有 SVD(比特征值分解更通用)
  2. 唯一性:奇异值唯一,奇异向量在符号意义下唯一
  3. 低秩近似:前 \(k\) 个奇异值给出最优秩-\(k\) 近似
  4. 条件数\(\kappa(\mathbf{A}) = \frac{\sigma_{\max}}{\sigma_{\min}}\) 表示矩阵的病态程度
  5. :$\text{rank}(\mathbf{A}) = $ 非零奇异值的个数
  6. 范数
  7. 谱范数:\(\|\mathbf{A}\|_2 = \sigma_{\max}\)
  8. Frobenius 范数:\(\|\mathbf{A}\|_F = \sqrt{\sum \sigma_i^2}\)

📡 在通信中的核心应用

1. MIMO 信道分解 ⭐⭐⭐

对于 MIMO 信道 \(\mathbf{y} = \mathbf{H}\mathbf{x} + \mathbf{n}\),对 \(\mathbf{H}\) 做 SVD: $$ \mathbf{H} = \mathbf{U}\mathbf{\Sigma}\mathbf{V}^H $$

预编码和解码: - 发送端预编码:\(\mathbf{x} = \mathbf{V}\mathbf{s}\) - 接收端解码:\(\tilde{\mathbf{y}} = \mathbf{U}^H\mathbf{y}\)

等效信道: $$ \tilde{\mathbf{y}} = \mathbf{U}^H\mathbf{H}\mathbf{V}\mathbf{s} + \mathbf{U}^H\mathbf{n} = \mathbf{\Sigma}\mathbf{s} + \tilde{\mathbf{n}} $$

物理意义: 将 MIMO 信道分解为 \(r\) 个并行的 SISO 子信道!

2. 波束成形(Beamforming)

  • 右奇异向量 \(\mathbf{V}\):最优发送波束成形矩阵
  • 左奇异向量 \(\mathbf{U}\):最优接收波束成形矩阵
  • 最大奇异值 \(\sigma_1\):最大信道增益方向

3. 信道压缩与反馈

  • 只反馈前 \(k\) 个奇异值和对应的奇异向量
  • 大幅减少反馈开销
  • 用于 FDD 系统的 CSI 反馈

4. 干扰抑制

  • 零空间波束成形:将信号投射到干扰信道的零空间
  • 利用 SVD 找到零空间基向量

3.5 矩阵求逆引理

🎯 Woodbury 矩阵恒等式

\[ (\mathbf{A} + \mathbf{UCV})^{-1} = \mathbf{A}^{-1} - \mathbf{A}^{-1}\mathbf{U}(\mathbf{C}^{-1} + \mathbf{V}\mathbf{A}^{-1}\mathbf{U})^{-1}\mathbf{V}\mathbf{A}^{-1} \]

其中: - \(\mathbf{A}\)\(n \times n\) 可逆矩阵 - \(\mathbf{C}\)\(k \times k\) 可逆矩阵 - \(\mathbf{U}\)\(n \times k\) 矩阵 - \(\mathbf{V}\)\(k \times n\) 矩阵

🔧 特例:Sherman-Morrison 公式

\(\mathbf{C} = 1\)(标量),\(\mathbf{U} = \mathbf{u}\)\(\mathbf{V} = \mathbf{v}^T\) 时: $$ (\mathbf{A} + \mathbf{u}\mathbf{v}T)} = \mathbf{A}^{-1} - \frac{\mathbf{A{-1}\mathbf{u}\mathbf{v}T\mathbf{A}^{-1}}{1 + \mathbf{v}T\mathbf{A} $$}\mathbf{u}

📡 在通信中的应用

1. 递归最小二乘(RLS)算法

在自适应滤波中,需要实时更新: $$ \mathbf{R}^{-1}[n] = (\mathbf{R}[n-1] + \mathbf{x}[n]\mathbf{x}H[n]) $$

使用 Sherman-Morrison 公式,避免每次重新求逆: $$ \mathbf{R}^{-1}[n] = \mathbf{R}^{-1}[n-1] - \frac{\mathbf{R}{-1}[n-1]\mathbf{x}[n]\mathbf{x}H[n]\mathbf{R}^{-1}[n-1]}{1 + \mathbf{x}H[n]\mathbf{R} $$}[n-1]\mathbf{x}[n]

计算复杂度:\(O(n^3)\) 降到 \(O(n^2)\)

2. MMSE 接收机

MMSE 滤波器: $$ \mathbf{W}_{\text{MMSE}} = (\mathbf{H}^H\mathbf{H} + \sigma2\mathbf{I})^H $$}\mathbf{H

当用户数变化时,使用矩阵求逆引理快速更新

3. 卡尔曼滤波

卡尔曼增益更新: $$ \mathbf{K}[n] = \mathbf{P}[n|n-1]\mathbf{H}T(\mathbf{H}\mathbf{P}[n|n-1]\mathbf{H}T + \mathbf{R})^{-1} $$

使用矩阵求逆引理简化计算


3.6 向量空间与投影

🎯 基本概念

线性组合

向量 \(\mathbf{v}\)\(\{\mathbf{v}_1, \mathbf{v}_2, \ldots, \mathbf{v}_k\}\) 的线性组合: $$ \mathbf{v} = c_1\mathbf{v}_1 + c_2\mathbf{v}_2 + \cdots + c_k\mathbf{v}_k $$

线性相关与无关

  • 线性相关:存在不全为零的 \(c_i\) 使得 \(\sum c_i\mathbf{v}_i = \mathbf{0}\)
  • 线性无关:只有 \(c_1 = c_2 = \cdots = c_k = 0\) 时才有 \(\sum c_i\mathbf{v}_i = \mathbf{0}\)

基与维度

  • :一组线性无关的向量,可以张成整个空间
  • 维度:基向量的个数

📐 正交与投影

内积(点积)

对于实向量: $$ \langle \mathbf{x}, \mathbf{y} \rangle = \mathbf{x}^T\mathbf{y} = \sum_{i=1}^{n} x_i y_i $$

对于复向量: $$ \langle \mathbf{x}, \mathbf{y} \rangle = \mathbf{x}^H\mathbf{y} = \sum_{i=1}^{n} \overline{x_i} y_i $$

正交

两个向量正交当且仅当内积为零: $$ \langle \mathbf{x}, \mathbf{y} \rangle = 0 $$

投影

向量 \(\mathbf{x}\) 在向量 \(\mathbf{y}\) 上的投影: $$ \text{proj}_{\mathbf{y}}(\mathbf{x}) = \frac{\langle \mathbf{x}, \mathbf{y} \rangle}{\langle \mathbf{y}, \mathbf{y} \rangle}\mathbf{y} $$

向量 \(\mathbf{x}\) 在子空间 \(\mathcal{S}\)(基矩阵 \(\mathbf{A}\))上的投影: $$ \text{proj}{\mathcal{S}}(\mathbf{x}) = \mathbf{A}(\mathbf{A}H\mathbf{A})}\mathbf{A}^H\mathbf{x} = \mathbf{P $$}}\mathbf{x

其中 \(\mathbf{P}_{\mathbf{A}} = \mathbf{A}(\mathbf{A}^H\mathbf{A})^{-1}\mathbf{A}^H\) 是**投影矩阵**

📡 在通信中的应用

1. 最小二乘估计

问题:求解 \(\mathbf{y} = \mathbf{H}\mathbf{x} + \mathbf{n}\) 中的 \(\mathbf{x}\)

最小二乘解: $$ \hat{\mathbf{x}} = (\mathbf{H}H\mathbf{H}) $$}\mathbf{H}^H\mathbf{y

几何解释\(\mathbf{H}\hat{\mathbf{x}}\)\(\mathbf{y}\)\(\mathbf{H}\) 列空间上的投影

2. 零强迫(ZF)接收机

\[ \mathbf{W}_{\text{ZF}} = (\mathbf{H}^H\mathbf{H})^{-1}\mathbf{H}^H \]

原理:将接收信号投影到信道的伪逆空间,消除用户间干扰

3. 信号检测

在存在干扰的情况下,将接收信号投影到期望信号子空间,抑制干扰


3.7 二次型与正定矩阵

🎯 二次型

定义

\[ q(\mathbf{x}) = \mathbf{x}^H\mathbf{A}\mathbf{x} = \sum_{i=1}^{n}\sum_{j=1}^{n} \overline{x_i}a_{ij}x_j \]

其中 \(\mathbf{A}\) 是 Hermitian 矩阵

正定性分类

类型 条件 特征值
正定 \(\mathbf{x}^H\mathbf{A}\mathbf{x} > 0, \forall \mathbf{x} \neq \mathbf{0}\) 全部 \(> 0\)
半正定 \(\mathbf{x}^H\mathbf{A}\mathbf{x} \geq 0, \forall \mathbf{x}\) 全部 \(\geq 0\)
负定 \(\mathbf{x}^H\mathbf{A}\mathbf{x} < 0, \forall \mathbf{x} \neq \mathbf{0}\) 全部 \(< 0\)
不定 可正可负 有正有负

📡 在通信中的应用

1. 信号能量

信号 \(\mathbf{x}\) 的能量: $$ E = \mathbf{x}^H\mathbf{x} = |\mathbf{x}|^2 $$

2. 协方差矩阵

随机向量 \(\mathbf{x}\) 的协方差矩阵: $$ \mathbf{R}_x = E[(\mathbf{x} - \boldsymbol{\mu})(\mathbf{x} - \boldsymbol{\mu})^H] $$

性质:总是半正定

3. 优化问题

在功率约束下最大化信噪比: $$ \max_{\mathbf{w}} \frac{\mathbf{w}H\mathbf{R}_s\mathbf{w}}{\mathbf{w}H\mathbf{R}_n\mathbf{w}} $$

这是广义瑞利商问题,最优解是广义特征向量


3.8 克拉默 - 拉奥下界(CRLB)中的线性代数

🎯 Fisher 信息矩阵

对于参数向量 \(\boldsymbol{\theta}\),Fisher 信息矩阵: $$ \mathbf{I}(\boldsymbol{\theta}) = E\left[\left(\frac{\partial \ln p(\mathbf{x};\boldsymbol{\theta})}{\partial \boldsymbol{\theta}}\right)\left(\frac{\partial \ln p(\mathbf{x};\boldsymbol{\theta})}{\partial \boldsymbol{\theta}}\right)^H\right] $$

📐 CRLB 定理

任何无偏估计量 \(\hat{\boldsymbol{\theta}}\) 的协方差矩阵满足: $$ \text{Cov}(\hat{\boldsymbol{\theta}}) \succeq \mathbf{I}(\boldsymbol{\theta})^{-1} $$

其中 \(\succeq\) 表示半正定序(即 \(\text{Cov} - \mathbf{I}^{-1}\) 是半正定矩阵)

📡 应用示例

信道估计的 CRLB:

对于 \(\mathbf{y} = \mathbf{H}\boldsymbol{\theta} + \mathbf{n}\)\(\mathbf{n} \sim \mathcal{CN}(\mathbf{0}, \sigma^2\mathbf{I})\)

Fisher 信息矩阵: $$ \mathbf{I}(\boldsymbol{\theta}) = \frac{1}{\sigma2}\mathbf{H}H\mathbf{H} $$

CRLB: $$ \text{Cov}(\hat{\boldsymbol{\theta}}) \succeq \sigma2(\mathbf{H}H\mathbf{H})^{-1} $$

LS 估计 \(\hat{\boldsymbol{\theta}}_{\text{LS}} = (\mathbf{H}^H\mathbf{H})^{-1}\mathbf{H}^H\mathbf{y}\) 达到 CRLB,是最优无偏估计!


📊 本章公式速查表

公式 表达式 应用场景
矩阵乘法 \((\mathbf{AB})_{ij} = \sum_k a_{ik}b_{kj}\) 所有 MIMO 运算
矩阵求逆 \(\mathbf{A}\mathbf{A}^{-1} = \mathbf{I}\) 信道均衡
Sherman-Morrison \((\mathbf{A}+\mathbf{u}\mathbf{v}^T)^{-1} = \mathbf{A}^{-1} - \frac{\mathbf{A}^{-1}\mathbf{u}\mathbf{v}^T\mathbf{A}^{-1}}{1+\mathbf{v}^T\mathbf{A}^{-1}\mathbf{u}}\) RLS 算法
特征值分解 \(\mathbf{A} = \mathbf{V}\mathbf{\Lambda}\mathbf{V}^{-1}\) PCA、功率分配
SVD \(\mathbf{A} = \mathbf{U}\mathbf{\Sigma}\mathbf{V}^H\) MIMO 预编码
投影矩阵 \(\mathbf{P}_{\mathbf{A}} = \mathbf{A}(\mathbf{A}^H\mathbf{A})^{-1}\mathbf{A}^H\) 最小二乘
MIMO 容量 \(C = \sum_i \log_2(1 + \frac{P_i\lambda_i}{\sigma^2})\) 系统性能分析
MMSE \(\mathbf{W} = (\mathbf{H}^H\mathbf{H} + \sigma^2\mathbf{I})^{-1}\mathbf{H}^H\) 接收机设计

🎯 本章习题

基础题

3.1 计算矩阵 \(\mathbf{A} = \begin{bmatrix} 2 & 1 \\ 1 & 2 \end{bmatrix}\) 的特征值和特征向量

3.2 对矩阵 \(\mathbf{B} = \begin{bmatrix} 1 & 0 \\ 0 & 2 \\ 0 & 0 \end{bmatrix}\) 进行 SVD 分解

3.3 使用 Sherman-Morrison 公式计算 \((\mathbf{I} + \mathbf{u}\mathbf{u}^H)^{-1}\),其中 \(\mathbf{u} = [1, 1]^T\)

应用题

3.4 对于 \(2 \times 2\) MIMO 信道 \(\mathbf{H} = \begin{bmatrix} 1 & 0.5 \\ 0.5 & 1 \end{bmatrix}\): - (a) 计算 \(\mathbf{H}\) 的 SVD - (b) 计算信道容量(假设 SNR = 10 dB) - © 设计最优预编码矩阵

3.5 推导 LS 估计量 \(\hat{\mathbf{x}} = (\mathbf{H}^H\mathbf{H})^{-1}\mathbf{H}^H\mathbf{y}\) 的协方差矩阵,并验证其达到 CRLB

编程题

3.6 使用 Python/NumPy 实现: - (a) 矩阵的特征值分解 - (b) SVD 分解 - © 比较与内置函数的结果


🔗 与后续章节的联系

后续章节 使用的线性代数知识
第 9 章 信道估计 最小二乘、MMSE、投影、CRLB
第 10 章 均衡算法 矩阵求逆、ZF、MMSE 接收机
第 11 章 MIMO SVD、特征值分解、波束成形
第 12 章 OFDM 酉矩阵、DFT 矩阵性质

💡 学习建议

  1. 动手推导:不要只看公式,亲自推导每个重要结论
  2. 几何直观:理解矩阵运算的几何意义(旋转、缩放、投影)
  3. 编程验证:用 NumPy/MATLAB 验证理论结果
  4. 联系应用:时刻思考"这在通信系统中对应什么?"
  5. 重点掌握:SVD、特征值分解、矩阵求逆引理是核心中的核心

第 3 章 完成 ✅
最后更新:2026-03-25