《《Information Theory & Coding信息论与编码(英文版)》-梁建武-电子教案 第6章 限失真编码》由会员分享,可在线阅读,更多相关《《Information Theory & Coding信息论与编码(英文版)》-梁建武-电子教案 第6章 限失真编码(66页珍藏版)》请在金锄头文库上搜索。
1、Chapter 6 Limited distortion Loss Source Coding,Brief introduction: compared with the lossless encoding,(comparison on distortion and entropy rate ). If the entropy compressed encoding permits distortion, then entropy rate will be compressed to the least after encoded. (without decoder)(explain the
2、necessity of two kinds of encode method) Info. rate distortion function R(D): is the basis of entropy compressed encoding. It combines the two measurements : information and distortion to provide solution to consider two factors at the same time in signal processing.,The necessity of introduction of
3、 distortion,1)Distortion is inevitable; 2)Whether the receiver (info. host) is human being or machine device, he has certain analyze ability and sensitivity. The transmission process surpass these ability will be no significance. 3)Even the receiver can be distinguished and be differentiated, it has
4、 little influence on communication quality. So we also call it permitted distortion; 4)Our goal is to research various objective source and source receiver and to get the biggest (permitted) distortion D and its least corresponding source info. rate R(D) when the Qos is assigned.,5)To limited distor
5、tion source, the least info. rate must be transmitted is R(D),not the source entropy H(U) under the lossless situation. Obviously, H(U)R(D). only when D=0,the equality is tenable; 6)We must establish objective distortion measure of the source in order to measure D,should we also establish stint rela
6、tion with D; 7)Function R(D) is the theory basis of limited distortion source processing;,Goals to achieve in this chapter,Understand the significance of limited distortion source encoding in communication system Discern the meaning and basic method of distortion measure Understand the concept, phys
7、ical significance and nature of function R(D) Discern the calculation method of function R(D) in specific situation Discern the usage of function R(D) in channel encoding theorem Understand the physical and practical significance of the Third Shannon theorem Discern the interrelation and comparison
8、of the three Shannon theorem,Contents of this chapter,outline Measure of loss Info. rate loss function Limited loss source coding theorem Relations and comparison of the three Shannon theorems,6.1:Outline1,Problem introduced Review of noiseless source coding theorem Review of noise source coding the
9、orem Problems existing Characteristics of practical demand Research content led into Research method and sequence of this chapter,6.1:Outline2,Review of noiseless source coding theorem We can inevitably find an input distribution (method of source coding) to transmit info. with channel capacity C an
10、d without error in noiseless lossless channel.,compress redundancy,use C best,6.1:Outline3,Review of noise source coding theorem So long as RC, we can find a method of source coding to transmit info. with PE as little as possible in the channel.,Source coding,channel R C; PE=,message,Channel coding,
11、increase redundancy, match the channel characteristics best,6.1:Outline4,Problems existing For continuous and analog source H(S)= Channel transmit rate R=H(S)/n(bit/code signal) R= Average code length l=Hr(S)=H(S)/logr, l= , In practical, because B is limited,C is definitely limited,RC and l= are im
12、possible,6.1:Outline5,Characteristics of practical demand Truth degree which source host demands Practical voice signal:20Hz8KHz our ears can tell apart:300Hz3400Hz Picture chromatism:can be enough visuognosis:256 bit(monochrome)is enough Certain infidelity is permitted Whole fidelity is unnecessary
13、,6.1:Outline6,Research content led into Problems of limited loss source encoding What degree will the source be compressed if certain distortion is permitted?(how much bits will need at least to describe source in the receive end?) How much will the most distortion be with certain info. transmit rat
14、e R? Relative problems How to measure distortion? How to calculate rate distortion function?,6.1:Outline7,Research method: Abstract channel Virtual test channel,6.1:Outline8,Method: Abstract:abstract the part which has slight relation with the discussion focus Because involving the source encoding,
15、abstract the channel Channel encode channel channel decode channel* channel* can be omitted According to the source encode theorem, channel* is a generalized channel with no disturbance, the info. distortion which the source host received only came from source encoding,6.1:Outline9,Method: virtual:m
16、ake the discussion focus virtual and detailed Make the encode process of limited distortion source virtual Channel encode channel channel decode test channel May use the channel transmission probability to describe the relation of not encoded and encoded limited distortion source,信源编码,信道*,信源译码,信源,信宿,信源,信宿,试验信道,U,V,P(V|U),6.1:Outline10,Discuss sequence of this chapter Begins from the simplest discrete non- memory so