最美情侣中文字幕电影,在线麻豆精品传媒,在线网站高清黄,久久黄色视频

歡迎光臨散文網(wǎng) 會(huì)員登陸 & 注冊

Turbo 譯碼淺析

2023-01-18 23:46 作者:樂吧的數(shù)學(xué)  | 我要投稿

本文章和系列視頻,將講解 Turbo 碼的基本入門。

需要的預(yù)備知識主要是知道卷積碼的基本編碼過程,知道卷積碼的 BCJR 譯碼算法。

Turbo 碼是由多個(gè)卷積碼并聯(lián)而成的, Turbo 名字的由來,是從譯碼的角度來看的。因此,嚴(yán)格來說,應(yīng)該稱為并聯(lián)卷積碼的 Turbo 譯碼算法更合適。
注:也有串聯(lián)的卷積碼構(gòu)成 Turbo 碼,也有由多個(gè)不同的卷積碼并聯(lián)/串聯(lián)成的 Turbo 碼,本文主要講解由多個(gè)相同的卷積碼并聯(lián)構(gòu)成的 Turbo 碼。而且,為了敘述方便和公式的簡潔,我們以下面的卷積碼為例子進(jìn)行討論。


如果單獨(dú)看每一個(gè)卷積碼,我們可以用 BCJR 譯碼算法來譯碼,假如我們先用上面那一個(gè)卷積碼,做 BCJR 譯碼,則根據(jù)之前的文章和視頻講解,我們知道對每個(gè)發(fā)送比特的后驗(yàn)概率,可以用如下公式計(jì)算:

%5Cbegin%7Baligned%7D%0AP(X_t%3Dx%7Cr)%20%26%3D%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7DP(%5Cpsi_t%3Dp%2C%5Cpsi_%7Bt%2B1%7D%3Dq%7Cr)%20%5C%5C%0A%5C%5C%0A%26%5Cpropto%20%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20%5Cgamma_t(p%2Cq)%20%5Cbeta_%7Bt%2B1%7D(q)%0A%5Cend%7Baligned%7D%20%20%5Ctag%201
其中:

%5Cbegin%7Baligned%7D%0A%5Cgamma_t(p%2Cq)%0A%26%3D%20%20p(%5Cpsi_%7Bt%2B1%7D%3Dq%2C%20r_t%20%7C%20%20%5Cpsi_t%3Dp)%20%5C%5C%20%20%5C%5C%0A%26%3D%20p(%20r_t%20%7C%20%5Cpsi_%7Bt%2B1%7D%3Dq%2C%20%20%5Cpsi_t%3Dp)%20p(%5Cpsi_%7Bt%2B1%7D%3Dq%7C%5Cpsi_t%3Dp)%20%20%5C%5C%20%20%5C%5C%0A%26%3D%20p(r_t%7Ca_t)%20p(x_t%3Dx)%0A%0A%0A%5Cend%7Baligned%7D%20%20%5Ctag%202


以及:

%5Cbegin%7Baligned%7D%0A%5Calpha_%7Bt%2B1%7D(q)%20%26%3D%5Csum_p%20%20%5Calpha_t(p)%20%5Clambda_t(p%2Cq)%20%20%20%5C%5C%20%20%5C%5C%0A%0A%5Cbeta_t(p)%20%26%3D%20%5Csum_q%20%20%5Cgamma_t(p%2Cq)%20%20%5Cbeta_%7Bt%2B1%7D(q)%0A%5Cend%7Baligned%7D%20%20%5Ctag%203

那么,我們可以很自然地想到,是否可以利用上面卷積碼得到的后驗(yàn)概率信息,來增強(qiáng)對第二個(gè)卷積碼做概率計(jì)算的可靠性?反過來,等第二個(gè)卷積碼的后驗(yàn)概率計(jì)算出來后,是否又可以用來增強(qiáng)對第一個(gè)卷積碼做概率計(jì)算的可靠性?如此往復(fù)迭代,就構(gòu)成了類似渦輪增壓的工作機(jī)制,因此,得名 Turbo 譯碼。

現(xiàn)在,我們來分析,從另外一個(gè)卷積碼的譯碼結(jié)果中,拿什么樣的概率信息給當(dāng)前這個(gè)卷積碼的譯碼使用。

最直觀的,最容易理解的,就是把第一個(gè)卷積碼的后驗(yàn)概率當(dāng)成對發(fā)送比特的概率,代入到第二個(gè)卷積碼中用到比特概率的地方。

如果令另外一個(gè)卷積碼中公式 (2) 里的先驗(yàn)概率?p(x_t%3Dx) 等于上一個(gè)卷積碼中的后驗(yàn)概率,即:
p(x_t%3Dx)%20%3D%20P(X_t%3Dx%7Cr)%20%20%5Ctag%204

我們來分析一下這樣做會(huì)有什么問題。從上面的公式,我們在計(jì)算?%5Cgamma 這個(gè)概率時(shí),用到了先驗(yàn)概率,我們把 %5Cgamma? 這個(gè)概率公式再展開分析一下,從公式 (2) 繼續(xù)分析。因?yàn)槲覀冇玫降木矸e碼是系統(tǒng)碼,即編碼的 bit 會(huì)在編碼后的碼流中原封不動(dòng)地保存,那么:
%5Cbegin%7Baligned%7D%0A%5Cgamma_t(p%2Cq)%0A%26%3D%20%20p(%5Cpsi_%7Bt%2B1%7D%3Dq%2C%20r_t%20%7C%20%20%5Cpsi_t%3Dp)%20%5C%5C%20%20%5C%5C%0A%26%3D%20p(%20r_t%20%7C%20%5Cpsi_%7Bt%2B1%7D%3Dq%2C%20%20%5Cpsi_t%3Dp)%20p(%5Cpsi_%7Bt%2B1%7D%3Dq%7C%5Cpsi_t%3Dp)%20%20%5C%5C%20%20%5C%5C%0A%26%3D%20p(r_t%5E%7B(0)%7D%2Cr_t%5E%7B(1)%7D%7Ca_t%5E%7B(0)%7D%2C%20a_t%5E%7B(1)%7D)%20p(x_t%3Dx)%0A%0A%0A%5Cend%7Baligned%7D%20%20%5Ctag%205

由于用的是系統(tǒng)碼,所以?a_t%5E%7B(0)%7D 對應(yīng)的就是 x_t,所以,公式 (5) 可以寫成:
%5Cbegin%7Baligned%7D%0A%5Cgamma_t(p%2Cq)%0A%26%3D%20p(r_t%5E%7B(0)%7D%2Cr_t%5E%7B(1)%7D%7Cx_t%2C%20a_t%5E%7B(1)%7D)%20p(x_t%3Dx)%20%20%20%5C%5C%0A%26%3D%20p(r_t%5E%7B(0)%7D%7Cx_t)p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20p(x_t%3Dx)%0A%0A%5Cend%7Baligned%7D%20%20%5Ctag%206把 (6)? 代入 (1) 則:
%5Cbegin%7Baligned%7D%0AP(X_t%3Dx%7Cr)%20%26%5Cpropto%20%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20%5Cgamma_t(p%2Cq)%20%5Cbeta_%7Bt%2B1%7D(q)%20%5C%5C%0A%26%20%5Cpropto%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(0)%7D%7Cx_t)p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20p(x_t%3Dx)%20%5Cbeta_%7Bt%2B1%7D(q)%5C%5C%0A%26%20%5Cpropto%20p(r_t%5E%7B(0)%7D%7Cx_t)p(x_t%3Dx)%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20%5Cbeta_%7Bt%2B1%7D(q)%0A%5Cend%7Baligned%7D%20%20%5Ctag%207

我們把公式 (7) 中最后三項(xiàng),分別記為:

%5Cbegin%7Baligned%7D%0AP_%7Bs%2Ct%7D(x)%20%26%3D%20p(r_t%5E%7B(0)%7D%7Cx_t)%20%5C%5C%20%5C%5C%0AP_%7Bp%2Ct%7D(x)%26%3Dp(x_t%3Dx)%20%5C%5C%20%5C%5C%0AP_%7Be%2Ct%7D(x)%26%3D%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20%5Cbeta_%7Bt%2B1%7D(q)%0A%5Cend%7Baligned%7D


如果把第一個(gè)卷積碼計(jì)算出來的后驗(yàn)概率,代入到第二個(gè)卷積碼的先驗(yàn)概率,則有:

%5Cbegin%7Baligned%7D%0AP(X_t%3Dx%7Cr)%20%26%20%5Cpropto%20p(r_t%5E%7B(0)%7D%7Cx_t)p(x_t%3Dx)%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(2)%7D%7C%20a_t%5E%7B(2)%7D)%20%5Cbeta_%7Bt%2B1%7D(q)%20%20%5C%5C%0A%26%20%5Cpropto%20%20p(r_t%5E%7B(0)%7D%7Cx_t)%20%20%20%20p(r_t%5E%7B(0)%7D%7Cx_t)p(x_t%3Dx)%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20%5Cbeta_%7Bt%2B1%7D(q)%20%20%20%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(2)%7D%7C%20a_t%5E%7B(2)%7D)%20%5Cbeta_%7Bt%2B1%7D(q)%0A%5Cend%7Baligned%7D%20%20%20%5Ctag%208

可以看到,上面有兩個(gè) p(r_t%5E%7B(0)%7D%7Cx_t),而且代入后,還有 p(x_t%3Dx),這樣會(huì)導(dǎo)致 “重復(fù)計(jì)算” 的感覺,參考書中好像是說不要這樣,而是用公式 (7) 中 求和的部分,作為第一個(gè)卷積碼因?yàn)榫幋a而得到的概率信息,給第二個(gè)卷積碼使用,即傳遞:

p(x_t%3Dx)%3D%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20%5Cbeta_%7Bt%2B1%7D(q)%20%20%5Ctag%209

所以, Turbo 譯碼的大致流程為:

1) 對第一個(gè)卷積碼,計(jì)算?%5Calpha%2C%20%5Cbeta 概率
2) 利用公式 (9),計(jì)算? P_%7Be%2Ct%7D(x)%3D%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(1)%7D%7C%20a_t%5E%7B(1)%7D)%20%5Cbeta_%7Bt%2B1%7D

3) 把?P_%7Be%2Ct%7D(x) 當(dāng)成 p(x_t%3Dx),傳遞給第二個(gè)卷積碼
4) 對第二個(gè)卷積碼,計(jì)算?%5Calpha%2C%20%5Cbeta 概率
5) 利用公式 (9) (稍微變化一點(diǎn)),計(jì)算? P_%7Be%2Ct%7D(x)%3D%5Csum_%7B(p%2Cq)%5Cin%20S_x%7D%20%5Calpha_t(p)%20p(r_t%5E%7B(2)%7D%7C%20a_t%5E%7B(2)%7D)%20%5Cbeta_%7Bt%2B1%7D

6) 把?P_%7Be%2Ct%7D(x) 當(dāng)成 p(x_t%3Dx),傳遞回第一個(gè)卷積碼,轉(zhuǎn)到步驟 1 繼續(xù),直到結(jié)束.


詳細(xì)的算法描述如下:

?j%5Cin%20%5C%7B1%2C2%5C%7D 表示第幾個(gè)卷積碼

l? 表示第幾次迭代

P_%7Be%2Ct%7D%5E%7B(l%2Cj)%7D,表示第 l 次迭代中,第?j ?個(gè)卷積碼計(jì)算出來的要傳遞的外信息。

P%5E%7B(l%2Cj)%7D? 表示第 l? 次迭代中,第 ?j ?個(gè)卷積碼用到的 先驗(yàn)概率

M? 最大迭代次數(shù)

-----------------------------算法------begin

初始化: 令?P%5E%7B(0%2C1)%7D(x_t%3Dx)%3DP%5E%7B(0)%7D(x_t%3Dx)? (用初始先驗(yàn)概率輸入給第一個(gè)卷積碼,一般是等概率分布的)

迭代: 按照迭代次數(shù)循環(huán) l%3D1%2C2%2C....M
?? 1. 用?P%5E%7B(l-1%2C1)%7D(x_t%3Dx) 作為先驗(yàn)概率?P(x_t%3Dx%20) 輸入給第一個(gè)卷積碼,計(jì)算
????? * 所有的 %5Calpha%2C%20%5Cbeta

????? * 計(jì)算 P_%7Be%2Ct%7D%5E%7B(l%2C1)%7D(x_t%3Dx)

?? 2. 令 P%5E%7B(l%2C2)%7D(x_t%3Dx)%20%3D%20%5Cprod%20%5BP_%7Be%2Ct%7D%5E%7B(l%2C1)%7D(x_t%3Dx)%5D

?? 3. 用?P%5E%7B(l%2C2)%7D(x_t%3Dx)%20 作為先驗(yàn)概率 P(x_t%3Dx%20),計(jì)算
????? * 所有的 %5Calpha%2C%20%5Cbeta

?? 4. 如果不是最后一次迭代
????? * 計(jì)算 P_%7Be%2Ct%7D%5E%7B(l%2Cl)%7D(x_t%3Dx)

????? * 令 P%5E%7B(l%2C1)%7D(x_t%3Dx)%20%3D%20%5Cprod%5E%7B-1%7D%20%5BP_%7Be%2Ct%7D%5E%7B(l%2C2)%7D(x_t%3Dx)%5D

?? 5. 否則,如果是最后一次迭代
????? * 用?P%5E%7B(l%2C2)%7D(x_t%3Dx)%20 作為先驗(yàn)概率 P(x_t%3Dx%20),計(jì)算 P(x_t%3Dx%7Cr)

????? * 解交織 P(x_t%3Dx%7Cr)%20%3D%20%5Cprod%5E%7B-1%7D%5BP(x_t%3Dx%7Cr)%5D
? ?

-----------------------------算法------end



我們以一個(gè)具體的例子來看譯碼的過程.


首先,做初始化,我們有的先驗(yàn)概率,是 0/1 比特的取值是等概率的,所以:
P%5E%7B(0%2C1)%7D(x_t%3D0)%3DP%5E%7B(0)%7D(x_t%3D0)%20%3D%20%5Cfrac%7B1%7D%7B2%7D%20%20%5C%5C%0AP%5E%7B(0%2C1)%7D(x_t%3D1)%3DP%5E%7B(0)%7D(x_t%3D1)%20%3D%20%5Cfrac%7B1%7D%7B2%7D

對于時(shí)刻 0 到時(shí)刻 9,我們用公式 (6) 計(jì)算出來?%5Cgamma 概率:
%5Cbegin%7Baligned%7D%0A%5Cgamma_0(p%3D0%2Cq%3D0)%20%26%3D%20p(r_t%5E%7B(0)%7D%7Cx_0%3D0)p(r_t%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D%3D-1)%20p(x_0%3D0)%20%20%5C%5C%0A%5Cgamma_0(p%3D0%2Cq%3D2)%20%26%3D%20p(r_t%5E%7B(0)%7D%7Cx_0%3D1)p(r_t%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D%3D%2B1)%20p(x_0%3D1)%20%5C%5C%0A...%5C%5C%0A%5Cgamma_0(p%3D3%2Cq%3D1)%20%26%3D%20p(r_t%5E%7B(0)%7D%7Cx_0%3D0)p(r_t%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D%3D-1)%20p(x_0%3D0)%20%5C%5C%0A%5Cgamma_0(p%3D3%2Cq%3D3)%20%26%3D%20p(r_t%5E%7B(0)%7D%7Cx_0%3D1)p(r_t%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D%3D%2B1)%20p(x_0%3D1)%20%5C%5C%0A%5Cend%7Baligned%7D


計(jì)算出所有的 %5Calpha
?初始化 0 時(shí)刻的 %5Calpha
%5Calpha_0(0)%20%3D%201%20%20%5C%5C%0A%5Calpha_0(1)%20%3D%200%20%5C%5C%0A%5Calpha_0(2)%20%3D%200%20%5C%5C%0A%5Calpha_0(3)%20%3D%200%0A
1 時(shí)刻的 %5Calpha

%5Calpha_1(0)%20%3D%5Csum_%7Bp%5Cin%5C%7B0%2C1%5C%7D%7D%20%20%5Calpha_0(p)%20%5Clambda_0(p%2C0)%20%3D%5Calpha_0(0)%20%5Clambda_0(0%2C0)%2B%5Calpha_0(1)%20%5Clambda_0(1%2C0)%20%5C%5C%0A%5Calpha_1(1)%20%3D%5Csum_%7Bp%5Cin%5C%7B2%2C3%5C%7D%7D%20%20%5Calpha_0(p)%20%5Clambda_0(p%2C1)%20%3D%5Calpha_0(2)%20%5Clambda_0(2%2C1)%2B%5Calpha_0(3)%20%5Clambda_0(3%2C1)%20%5C%5C%0A%5Calpha_1(2)%20%3D%5Csum_%7Bp%5Cin%5C%7B0%2C1%5C%7D%7D%20%20%5Calpha_0(p)%20%5Clambda_0(p%2C2)%20%3D%5Calpha_0(0)%20%5Clambda_0(0%2C2)%2B%5Calpha_0(1)%20%5Clambda_0(1%2C2)%20%5C%5C%0A%5Calpha_1(3)%20%3D%5Csum_%7Bp%5Cin%5C%7B2%2C3%5C%7D%7D%20%20%5Calpha_0(p)%20%5Clambda_0(p%2C3)%20%3D%5Calpha_0(2)%20%5Clambda_0(2%2C3)%2B%5Calpha_0(3)%20%5Clambda_0(3%2C3)%20%5C%5C



依次類推,最后計(jì)算 9 時(shí)刻的 %5Calpha

%5Calpha_9(0)%20%3D%5Csum_%7Bp%5Cin%5C%7B0%2C1%5C%7D%7D%20%20%5Calpha_8(p)%20%5Clambda_8(p%2C0)%20%3D%5Calpha_8(0)%20%5Clambda_8(0%2C0)%2B%5Calpha_8(1)%20%5Clambda_8(1%2C0)%20%5C%5C%0A%5Calpha_9(1)%20%3D%5Csum_%7Bp%5Cin%5C%7B2%2C3%5C%7D%7D%20%20%5Calpha_8(p)%20%5Clambda_8(p%2C1)%20%3D%5Calpha_8(2)%20%5Clambda_8(2%2C1)%2B%5Calpha_8(3)%20%5Clambda_8(3%2C1)%20%5C%5C%0A%5Calpha_9(2)%20%3D%5Csum_%7Bp%5Cin%5C%7B0%2C1%5C%7D%7D%20%20%5Calpha_8(p)%20%5Clambda_8(p%2C2)%20%3D%5Calpha_8(0)%20%5Clambda_8(0%2C2)%2B%5Calpha_8(1)%20%5Clambda_8(1%2C2)%20%5C%5C%0A%5Calpha_9(3)%20%3D%5Csum_%7Bp%5Cin%5C%7B2%2C3%5C%7D%7D%20%20%5Calpha_8(p)%20%5Clambda_8(p%2C3)%20%3D%5Calpha_8(2)%20%5Clambda_8(2%2C3)%2B%5Calpha_8(3)%20%5Clambda_8(3%2C3)%20%5C%5C

再計(jì)算?%5Cbeta 概率,計(jì)算時(shí)刻 10 到時(shí)刻 1 的:

初始化時(shí)刻 10 的 ?%5Cbeta

%5Cbeta_%7B10%7D(0)%20%3D%201%20%20%5C%5C%0A%5Cbeta_%7B10%7D(1)%20%3D%200%20%5C%5C%0A%5Cbeta_%7B10%7D(2)%20%3D%200%20%5C%5C%0A%5Cbeta_%7B10%7D(3)%20%3D%200


計(jì)算 9 時(shí)刻的 ?%5Cbeta

%5Cbeta_9(0)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B0%2C2%5C%7D%7D%20%20%5Cgamma_9(0%2Cq)%20%20%5Cbeta_%7B10%7D(q)%20%3D%20%5Cgamma_9(0%2C0)%20%20%5Cbeta_%7B10%7D(0)%20%2B%20%20%5Cgamma_9(0%2C2)%20%20%5Cbeta_%7B10%7D(2)%20%5C%5C%0A%5Cbeta_9(1)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B0%2C2%5C%7D%7D%20%20%5Cgamma_9(1%2Cq)%20%20%5Cbeta_%7B10%7D(q)%20%3D%20%5Cgamma_9(1%2C0)%20%20%5Cbeta_%7B10%7D(0)%20%2B%20%20%5Cgamma_9(1%2C2)%20%20%5Cbeta_%7B10%7D(2)%20%5C%5C%0A%5Cbeta_9(2)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B1%2C3%5C%7D%7D%20%20%5Cgamma_9(2%2Cq)%20%20%5Cbeta_%7B10%7D(q)%20%3D%20%5Cgamma_9(2%2C1)%20%20%5Cbeta_%7B10%7D(1)%20%2B%20%20%5Cgamma_9(2%2C3)%20%20%5Cbeta_%7B10%7D(3)%20%5C%5C%0A%5Cbeta_9(3)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B1%2C3%5C%7D%7D%20%20%5Cgamma_9(3%2Cq)%20%20%5Cbeta_%7B10%7D(q)%20%3D%20%5Cgamma_9(3%2C1)%20%20%5Cbeta_%7B10%7D(1)%20%2B%20%20%5Cgamma_9(3%2C3)%20%20%5Cbeta_%7B10%7D(3)%20%5C%5C

以此類推,計(jì)算 1 時(shí)刻的 ?%5Cbeta

%5Cbeta_1(0)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B0%2C2%5C%7D%7D%20%20%5Cgamma_1(0%2Cq)%20%20%5Cbeta_2(q)%20%3D%20%5Cgamma_1(0%2C0)%20%20%5Cbeta_2(0)%20%2B%20%20%5Cgamma_1(0%2C2)%20%20%5Cbeta_2(2)%20%5C%5C%0A%5Cbeta_1(1)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B0%2C2%5C%7D%7D%20%20%5Cgamma_1(1%2Cq)%20%20%5Cbeta_2(q)%20%3D%20%5Cgamma_1(1%2C0)%20%20%5Cbeta_2(0)%20%2B%20%20%5Cgamma_1(1%2C2)%20%20%5Cbeta_2(2)%20%5C%5C%0A%5Cbeta_1(2)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B1%2C3%5C%7D%7D%20%20%5Cgamma_1(2%2Cq)%20%20%5Cbeta_2(q)%20%3D%20%5Cgamma_1(2%2C1)%20%20%5Cbeta_2(1)%20%2B%20%20%5Cgamma_1(2%2C3)%20%20%5Cbeta_2(3)%20%5C%5C%0A%5Cbeta_1(3)%20%3D%20%5Csum_%7Bq%5Cin%20%5C%7B1%2C3%5C%7D%7D%20%20%5Cgamma_1(3%2Cq)%20%20%5Cbeta_2(q)%20%3D%20%5Cgamma_1(3%2C1)%20%20%5Cbeta_2(1)%20%2B%20%20%5Cgamma_1(3%2C3)%20%20%5Cbeta_2(3)%20%5C%5C


至此,我們可以計(jì)算 extrinsic probability 外信息:



%5Cbegin%7Baligned%7D%0AP_%7Be%2C0%7D(0)%26%3D%5Csum_%7B(p%2Cq)%5Cin%20S_0%7D%20%5Calpha_0(p)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(q)%20%5C%5C%0A%26%3D%20%5Calpha_0(0)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(0)%20%2B%20%5Calpha_0(1)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(2)%20%20%5C%5C%0A%26%5Cquad%20%5Cquad%20%2B%5Calpha_0(2)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(1)%20%2B%20%5Calpha_0(3)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(3)%20%5C%5C%0A%0AP_%7Be%2C0%7D(1)%26%3D%5Csum_%7B(p%2Cq)%5Cin%20S_1%7D%20%5Calpha_0(p)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(q)%20%5C%5C%0A%26%3D%20%5Calpha_0(0)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(2)%20%2B%20%5Calpha_0(1)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(0)%20%20%5C%5C%0A%26%5Cquad%20%5Cquad%20%2B%5Calpha_0(2)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(3)%20%2B%20%5Calpha_0(3)%20p(r_0%5E%7B(1)%7D%7C%20a_0%5E%7B(1)%7D)%20%5Cbeta_%7B1%7D(1)%20%5C%5C%20%20%5C%5C%0A%0A%5Cend%7Baligned%7D



我們把下一個(gè)卷積碼要用的先驗(yàn)概率用前面的外信息賦值:
P%5E%7B(0%2C2)%7D(x_t%3D0)%3DP_%7Be%2C0%7D(0)%20%5C%5C%0AP%5E%7B(0%2C2)%7D(x_t%3D1)%3DP_%7Be%2C0%7D(1)用同樣的步驟計(jì)算 %5Cgamma%2C%20%5Calpha%2C%20%5Cbeta, 然后計(jì)算出?P_%7Be%2C0%7D(x) 外信息,把這個(gè)外信息作為下一輪的第一個(gè)卷積碼的先驗(yàn)概率來使用,繼續(xù)上面的過程。

Turbo 譯碼淺析的評論 (共 條)

分享到微博請遵守國家法律
南涧| 乌鲁木齐市| 疏勒县| 资溪县| 绥化市| 吴江市| 太仓市| 个旧市| 临高县| 安岳县| 广州市| 合江县| 南江县| 小金县| 沾化县| 泉州市| 宁河县| 贵溪市| 白沙| 金川县| 乃东县| 社旗县| 兴隆县| 建阳市| 钦州市| 沧源| 邻水| 泰州市| 循化| 荔波县| 汶上县| 馆陶县| 若羌县| 鸡泽县| 桐柏县| 射阳县| 丰县| 乌鲁木齐市| 防城港市| 汶上县| 社会|