ℹ

Block Codes for the Gaussian Channel

Definition (Block Codes for the Gaussian Channel)

Given positive integers nn and M=MnM=M_{n}, a block (fixed-length) code Cn=(n,M)\mathcal{C}_{n}=(n,M) for a AWGN Channel with

  • Average Power Constraint: PP
  • Codeword Length: nn
  • Number of Codewords: MM
  • Rate: 1nlog⁡2M\frac{1}{n}\log_{2}M (bits per channel symbol)
  • Message Set: M={1,2,…,M}\mathcal{M}=\{1,2,\dots,M\} of MM information messages intended for transmission (note: messages are uniformly distributed over M\mathcal{M})
  • Encoding Function: f:M→Rnf:\mathcal{M}\to\mathbb{R}^{n}yielding codewords c1=f(1),…,cM=f(M)c_{1}=f(1),\ldots,c_{M}=f(M) such that each codeword cm=(cm1,…,cmn)c_{m}=(c_{m_{1}},\ldots,c_{m_{n}}) of length nn satisfies the average power constraint PP 1n∑i=1ncmi2≤P, m=1,…,M\frac{1}{n}\sum\limits_{i=1}^{n}c_{m_{i}}^{2}\le P, \ m=1,\ldots,M
  • Decoding function g:Rn→Mg:\mathbb{R}^{n}\to\mathcal{M}

Lemma (Code Reliability (Continuous))

Pe(Cn)=1M∑m=1Mλm(Cn)P_{e}(\mathcal{C}_{n})= \frac{1}{M}\sum\limits_{m=1}^{M}\lambda_{m}(\mathcal{C}_{n})where λm(Cn)=P(g(Yn)≠m∣Xn=cm)=∫yn∈Rn: g(yn)≠mfYn∣Xn(yn∣cm)dyn\begin{align*} \lambda_{m}(\mathcal{C_{n}})&=P(g(Y^{n})\not=m|X^{n}=c_{m})\\ &=\int_{y^{n}\in\mathbb{R}^{n}: \ g(y^{n})\not=m}f_{Y^n|X^{n}}(y^{n}|c_{m})dy^{n} \end{align*}is the conditional error probability given that message m∈Mm\in\mathcal{M} is sent over the channel (via codeword cm=f(m)∈Rnc_{m}=f(m)\in\mathbb{R}^{n}).

Linked from