自适应多描述编码应对混合网络动态PLR与BER挑战

168 浏览量更新于2024-08-26 收藏 575KB PDF 举报

"这篇研究论文探讨了在具有动态包丢失率(PLR)和误比特率(BER)的混合网络中应用自适应多描述编码(AMDC)的技术。作者Meng Yang, Xuguang Lan和Nanning Zheng来自西安交通大学的人工智能与机器人研究所。文章在2012年12月首次提交，经过修订后于2013年4月再次提交，并于同年6月被接受，最终于7月在线发布。关键词包括多描述编码、索引分配、误比特率、包丢失率和混合网络。" 正文: 这篇论文主要关注的是在混合网络环境下，如何通过自适应多描述编码来同时解决包丢失和误比特率问题。在混合网络中，数据传输可能会受到两种主要的干扰因素：包丢失和传输过程中的错误，这会严重影响信号的传输性能。当PLR和BER随着时间动态变化时，这个问题变得更加复杂。传统解决方案通常侧重于单一问题，要么是包丢失问题，要么是误比特问题。然而，该论文提出了一种创新的AMDC方案，结合了索引分配(IA)技术，旨在同时解决这两个问题。AMDC的基本思想是在编码过程中增加冗余信息，以补偿可能丢失的数据包。通过这种方式，即使在网络条件不稳定时，也能保证一定程度的数据完整性。关键在于，论文中提出的AMDC方案能精确控制冗余率，以便根据给定的PLR优化端到端的传输性能。这意味着系统能够根据网络状况实时调整编码策略，以最大程度地减少由于包丢失和误比特导致的失真。这种自适应性对于提高服务质量(QoS)和用户体验质量(QoE)至关重要。此外，索引分配技术在AMDC中扮演着重要角色，它决定了不同描述之间的优先级和恢复顺序。在包丢失或错误发生时，适当的索引分配可以帮助接收端优先恢复关键信息，从而降低整体传输错误的影响。这篇论文为混合网络环境下的高效、鲁棒的数据传输提供了一种新的解决方案。AMDC和索引分配的结合展示了在动态网络条件下优化传输性能和容错能力的潜力，对于未来通信系统的可靠性和效率提升具有重要意义。

Simulating results are shown in Section 4 and the conclu-

sion of this work is in Section 5.

2. Problem formulation

In this section, we firstly solve the decoding problem in

Fig. 1(a), because it is the basis of the whole framework.

2.1. Decoding analysis

Let X⊂ℝ be a discrete alphabet source. f

ðxÞ is the

probability density function (PDF). We consider each

channel of the network as a binary error memoryless

channel (BEC). It has the input alphabet f0; 1g and the

output may be intact or inverted. We denote e∈f0; 1g as the

error. The probabilities of error transition are supposed as

Pð1j0Þ¼Pð0j1Þ¼p. Supposing the input sequence as

X ¼ x

; x

; :::, each of the elements is quantized by

scalar quantizer and bijective mapped to a finite central

codebook C by the function Q

: X-C ¼f0; 1; 2; :::; M1g

with M elements. Then in AMDQ step, each central code-

word is mapped to a side codeword pair to generate two

descriptions by the function Q

: C-fI; Jg, where

I≜f I

; I

; :::; I

1

g and J≜fJ

; J

; :::; J

1

g are two finite side

codebooks. Each element in set I and J can be rewritten as

a binary format: i

:::i

1

∈f0; 1g

and j

:::j

1

∈f0; 1g

where b

¼ ⌈log

⌉ and b

¼ ⌈log

⌉. The bit rates of

two side descriptions and the central description of MDC

are R

¼ log

, R

¼ log

, and R

¼ log

M (before

entropy coding) respectively, and the redundancy rate of

MDC, denoted γ, can be calculated as

γ ¼

þ R

log

ð1Þ

We denote the generated two sequences of the input X

as I

and J

. They are entropy coded by variable length

code (VLC), and transmitted in a BEC channel separately. At

the terminal, the received two sequences are decoded and

denoted

and

. Then the signal is reconstructed as

Y ¼ y

; y

; :::. The optimization problem is to find the

reconstructed signal

X of the original input X, that is

X ¼ argMAX

½PðXjYÞ ð2Þ

We suppose the input sequence as an independent and

identical distributed (i.i.d.) source. For the correlated

source, such as Markovian source, it can be learnt from

[17,20,28]. Then (2) can be rewritten as

X ¼ argMAX

∏

i ¼ 0; 1;:::

Pðx

¼ argMAX

∏

i ¼ 0; 1;:::

PðI

; J

;

¼ argMAX

∏

i ¼ 0; 1;:::

ðPðI

ÞPðJ

ÞÞ

ð3Þ

PðI

Þ is the transiting probability from

to I

, and can be

calculated as

PðI

Þ¼p

hðI

ð1pÞ

hðI

ð4Þ

where hðÞ denotes the hamming distance function. Simi-

larly, PðJ

Þ can be calculated as

PðJ

Þ¼p

hðJ

ð1pÞ

ðJ

ð5Þ

It is obvious that

X is related with AMDQ and p.Thegoal

is to find an optimal AMDQ solution towards the known

BER of the network. Combining all the elements of I and J,

we can get N

possible pairs, which can be equivalently

denoted as following combined binary format:

W ¼fw

¼ i

:::i

1

:::j

1

∈f0; 1g

þb

;

m ¼ 0; 1; :::; N

1g: ð6Þ

M elements in W will be used to bijective map with the

elements in the set C,denotedV

¼fv

∈W; m ¼ 1;

2; :::; Mg.From(1) we know that, M is related with the

redundancy rate of MDC. If the transmission is error-free, all

possible set V

are equivalent. That is because AMDQ is

reversible. Otherwise, they are not equivalent, because of

the bit errors over the wireless channel. Different set V

has different bit-error resilience for the system.

Take the IA of Fig. 2(b) for example, where all the block s

stand for the set W and the blocks with indices stand for

the set V

. We assume the input as a single sample x ¼ 11.

It is mapped to the pair ð2; 3Þ as two side descriptions,

whose binary format is ð

010 011

Þ. When 1-bit error

occurs, there are 6 possible cases for the received code-

word, that is (2,1), (2,2), (3,3), (6,3), (0,3), and (2,7), whose

1 2 3 4 5 N

+1 N

+2 N

+3 N

+5 2N

-1)

12345

…

0 1 3

01234567

2 5 6 8

4 7 10 11 13

9 12 15 16

14 18 19 21

17 20 23 24 26

22 25 28 29

27 30 31

Fig. 2. Index table. (a) Initialized index table. (b) MDSQ (M¼32).

M. Yang et al. / Signal Processing: Image Communication 28 (2013) 1132–114 21134

剩余10页未读，继续阅读

weixin_38693419

粉丝: 6
资源: 931

自适应多描述编码应对混合网络动态PLR与BER挑战

PLR.rar_PLR分段_PLR表示_matlab滑窗_滑窗方法matlab_自顶向下分段

AOS通信系统中基于最小丢包率的自适应调制编码方法

matlab plr模型直线回归将时间序列数据表示成相邻的线段簇

plr模型 python

迁移学习matlab PLR

请生成一段c语言，输入长度N，信源01出现的概率，二进制对称信道BSC的差错概率，输出不同差错概率下的误码率丢包率

使用 packet_loss_rate 函数计算从源节点和中继节点发送到目标节点的数据包的丢包率

简述误码率和丢包率的定义和区别

function plr = packet_loss_rate(power, distances) signal_strength = power ./ (distances .^ 2); threshold = 0.25; % 假设接收成功的信号强度阈值 received_packets = sum(signal_strength > threshold, 2); plr = (size(distances, 1) - received_packets) ./ size(distances, 1); end

最新资源