370 likes | 588 Views
Paper Presentation Channel Coding and Transmission Aspects for Wireless Multimedia. Authors: Joachim Hagenauer, Thomas Stochhammer Source: Proceedings of the IEEE , Volume: 87 Issue: 10 , Oct 1999, pp. 1764 -1777 Originally Presented by Hong Hong Chang, Feb 17, 2003. Overview. Introduction
E N D
Paper PresentationChannel Coding and Transmission Aspects for Wireless Multimedia Authors: Joachim Hagenauer, Thomas Stochhammer Source: Proceedings of the IEEE , Volume: 87 Issue: 10 , Oct 1999, pp. 1764 -1777 Originally Presented by Hong Hong Chang, Feb 17, 2003
Overview • Introduction • System Architecture • The Links between Source and Channel Coding • RCPC, UEP • PCM Transmission example • Transmission (C) 2005 by Yu Hen Hu
Wireless Channel • Multipath fading • Doppler spreading • Effect of distance • Quite noisy • High BER • average error rates up to 10% • Channel coding is necessary http://www.wireless.per.nl:202/multimed/cdrom97/indoor.htm (C) 2005 by Yu Hen Hu
Source Coding & Channel Coding (I) • Shannon’s separation theorem • source coding - long blocks of source symbols • channel coding -a sequence of random block codes with infinite length • Infinite delay data Source Coding Channel Coding Modulation transmission (C) 2005 by Yu Hen Hu
Source Coding & Channel Coding (II) • Shannon’s separation theorem is no longer applicable • short blocks, small delays • Combined and joint source and channel coding • MPEG II audio layer • Source-controlled channel decoding • Uses the residual redundancy of the uncompressed or partly compressed source data to improve channel decoding (C) 2005 by Yu Hen Hu
Transmissions - Two Kinds of Data Channels • Mode 1 • Error free delivery • Using ARQ • Delay and bit throughput rate (BTR) vary according to the channel conditions • Mode 2 • Guarantees constant bit rate and delay • Errors occur (C) 2005 by Yu Hen Hu
System for Transmission of Multimedia Applications over Mobile Channels S C M A C M A M A (C) 2005 by Yu Hen Hu
Application Properties • Delay-sensitive applications • Speech, video telephony • Use frequent resynchronization, reduced predictive coding • No ARQ, deep interleaving or long block codes • BTR-sensitive applications • Audio, video • Use bidirectional predictive coding, long term rate control algorithms • Might use error protection interleaving, serial or parallel concatenated coding or ARQ to exploit the provided bandwidth as optimally as possible (C) 2005 by Yu Hen Hu
Application Properties (Cont) • BER-sensitive applications • Data • Error-free delivery • Use ARQ, FEC (C) 2005 by Yu Hen Hu
Multimedia Transmission • Each application may request different QoS • All application are combined into one single transmission stream • New layer necessary for multimedia transmission Adaptation Layer Multiplex Layer (C) 2005 by Yu Hen Hu
Adaptation Layer and Multiplex Layer • Adaptation layer • Adapt the requesting upper application to transmission condition according to the required QoS • Have tools for error detection, error correction, bit reordering, retransmission protocols • Multiplex layer • Multiplex the adaptation layer bit streams or packets into one single bit steam • Optimizing the throughput, minimize misdeliveries (C) 2005 by Yu Hen Hu
Transmission Scheme over a Mobile Channel (C) 2005 by Yu Hen Hu
Links between Source Coding and Channel Coding • Channel State Information (CSI) • Connected by soft decision of demodulator/detector • Soft decision gains 2-3dB • Source Significant Information (SSI) • For unequal error protection (UEP) • Rate-compatible punctured convolutional code (RCPC) • Decision Reliability Information (DRI) • Soft output from channel decoder • Source a priori/a posteriori information (SAI) • probability of next bit, correlation • Reduce channel decoder error rate (C) 2005 by Yu Hen Hu
Rate-Compatible Punctured Convolutional Code for Unequal Error Protection • Start with a rate 1/n0linear convolutional code • Encode k input bits to produce n0k output bits • Delete n0k−n bits from the output bits • The code rate is • The corresponding n0k perforation matrix has n ones and n0k−n zeros http://www.ee.byu.edu/ee/class/ee685/lectures/lecture37.pdf (C) 2005 by Yu Hen Hu
Punctured Convolutional Code Example http://www.ee.byu.edu/ee/class/ee685/lectures/lecture37.pdf (C) 2005 by Yu Hen Hu
Puncture Pattern and Perforation Matrix http://www.ee.byu.edu/ee/class/ee685/lectures/lecture37.pdf (C) 2005 by Yu Hen Hu
Rate Compatible Convolutional Code 2/3 2/3 http://www.ee.byu.edu/ee/class/ee685/lectures/lecture37.pdf (C) 2005 by Yu Hen Hu
Rate Compatible Punctured Convolutional Code • A family of punctured codes are rate compatibleif the codeword bits from the higher-rate code are embedded in the lower rate codes. • The zeros in perforation matrices of the lower rate codes are also the zeros in the perforation matrices of the higher rate • The ones in in perforation matrices of the higher rate codes are also ones in in perforation matrices of the lower rate codes. http://www.ee.byu.edu/ee/class/ee685/lectures/lecture37.pdf (C) 2005 by Yu Hen Hu
RCPC Example http://www.ee.byu.edu/ee/class/ee685/lectures/lecture37.pdf (C) 2005 by Yu Hen Hu
Recursive Systematic Encoder Structure • Memory M=4 , Mother code rate = ½, Puncturing rate = 8/12 • Nonsystematic vs Systematic G(D) = (1+D3+D4, 1+D+D2+D4, 1+D2+D3+D4) Gs(D) = (C) 2005 by Yu Hen Hu
Error Probability Upper Bound • df– free distance, the minimum distance of any path from the correct path • cd– the sum of all information weights on all wrong path of distance d starting inside one puncturing period • Pd – the pairwise error probability of two code sequences at distance d (C) 2005 by Yu Hen Hu
Puncturing Table (C) 2005 by Yu Hen Hu
Comparison of systematic recursive convolutional code with nonsystematic codes (C) 2005 by Yu Hen Hu
Encoder & Decoder • Encoder • Puncture • Repeat – replacing “1” by “2” or any higher integer in the puncturing tables • Decoder • Punctured bits are stuffed with zeros • Repeated bits are combined by adding soft values • Header of frame contains the coding rate information of payload • Easily adapted to multimedia and channel requirements via puncturing control (C) 2005 by Yu Hen Hu
BER Performance of Systematic Recursive PCPC code (C) 2005 by Yu Hen Hu
Soft-In/Soft-Out Decoding • Decoding algorithm • Viterbi (VA) • Maximum-a-posteriori-probability-symbol-by-symbol (MAP) • VA and MAP can accept soft values • Source a priori information • Channel state information • VA and MAP can deliver soft outputs (C) 2005 by Yu Hen Hu
PCM Transmission example - EEP • Analog source • Source coding: m-bit linear quantization (m=20) • Quantized sample • smaller k -> more important. • Transmission distortion • equal Pbfor all k=1,2,…,m (C) 2005 by Yu Hen Hu
PCM Transmission Example – Applying Soft Bits • CSI is transformed to a DRI and directly passed to the source decoder. Thus, λ(x) (soft value) is obtained • Reconstructed PCM value • Gain of about 1.6dB in SNRPCM (C) 2005 by Yu Hen Hu
PCM Transmission Example – Apply Channel Coding • m=10 • m is smaller, quantization noise increases • Channel coding rate = ½ • RCPSRC 8/16 • Improves total performance (C) 2005 by Yu Hen Hu
PCM Transmission Example – UEP • Let all bits contribute the same transmission distortion. Then, • Small k, small Pb • Use this information for unequal error-protection design • Require that transmission distortion of each bit is smaller than quantization distortion. We have (C) 2005 by Yu Hen Hu
PCM Transmission Example: RCPSRC code for UEP • Employ • the upper bound for the bit error probability • Distance spectra of puncture table • Obtain a certain rate R(k) for each bit class at different channel SNR • Rate distribution for PCM Transmission (C) 2005 by Yu Hen Hu
PCM Transmission Example - Simulation Results (C) 2005 by Yu Hen Hu
Approaches to Improve the Transmission of Multimedia I. Error Resilient Source Coding • Fixed length coding • more stable against channel error • MPEG-4 error resilient mode • Space the Resync markers evenly throughout the bit stream • All predictively encoded information is confined within one video packet to prevent the propagation of errors (C) 2005 by Yu Hen Hu
II. Improved Receiver Algorithms • European Digital Satellite TV-Broadcasting standard • MPEG-2 based source coding • Concatenated coding scheme • Error-concealment techniques based on temporal, spatial, frequency • Joint-source channel coding • Instead of remove residual redundancy by using VLC, keep it and use it at the receiver side to achieve more reliable decoding • Soft source decoding (C) 2005 by Yu Hen Hu
III. Source Adapted UEP • RCPC • Application to GSM speech • Turbo Code • Channel coding is applied according to the bit sensitivity • Application to hierarchical video broadcast • Base layer and enhancement layer (C) 2005 by Yu Hen Hu
IV. Channel Adapted Combined Source-Channel Coding Methods • Goal • Allocate bit rates in an optimal way between source and channel encoders as the source and channel vary • Minimize end-to-end distortion • Feed back the CSI from the decoder to the encoder on a reverse channel (C) 2005 by Yu Hen Hu