[Paper Review] On the Joint Source-Channel Coding Error Exponent for Discrete Memoryless Systems: Computation and Comparison with Separate Coding

Yangfan Zhong, Fady Alajaji+1 more|ArXiv.org|Jan 4, 2006

Wireless Communication Security Techniques51 references3 citations

TL;DR

This paper computes and compares the joint source-channel coding (JSCC) error exponent $E_J$ with the tandem (separate) coding error exponent $E_T$ for discrete memoryless systems. Using Csiszár's bounds and Arimoto's algorithm, it shows $E_T \leq E_J \leq 2E_T$, with $E_J$ often close to $2E_T$, implying a power gain exceeding 2 dB in practical systems like binary sources over AWGN and Rayleigh fading channels with finite quantization.

ABSTRACT

We investigate the computation of Csiszar's bounds for the joint source-channel coding (JSCC) error exponent, E_J, of a communication system consisting of a discrete memoryless source and a discrete memoryless channel. We provide equivalent expressions for these bounds and derive explicit formulas for the rates where the bounds are attained. These equivalent representations can be readily computed for arbitrary source-channel pairs via Arimoto's algorithm. When the channel's distribution satisfies a symmetry property, the bounds admit closed-form parametric expressions. We then use our results to provide a systematic comparison between the JSCC error exponent E_J and the tandem coding error exponent E_T, which applies if the source and channel are separately coded. It is shown that E_T <= E_J <= 2E_T. We establish conditions for which E_J > E_T and for which E_J = 2E_T. Numerical examples indicate that E_J is close to 2E_T for many source-channel pairs. This gain translates into a power saving larger than 2 dB for a binary source transmitted over additive white Gaussian noise channels and Rayleigh fading channels with finite output quantization. Finally, we study the computation of the lossy JSCC error exponent under the Hamming distortion measure.

Motivation & Objective

• To compute the joint source-channel coding (JSCC) error exponent $E_J$ for discrete memoryless sources and channels.
• To derive explicit formulas for the rates at which Csiszár’s bounds on $E_J$ are attained.
• To systematically compare $E_J$ with the tandem coding error exponent $E_T$ in terms of performance and reliability.
• To quantify the performance gain of JSCC over separate coding in terms of error exponent and power efficiency.
• To extend the analysis to the lossy JSCC case under Hamming distortion.

Proposed method

• Derives equivalent expressions for Csiszár’s bounds on the JSCC error exponent $E_J$ using Fenchel’s duality and parametric representations.
• Applies Arimoto’s algorithm to compute $E_J$ efficiently for arbitrary source-channel pairs.
• Establishes closed-form expressions for $E_J$ when the channel exhibits symmetry, simplifying computation.
• Uses the random-coding exponent and sphere-packing exponent frameworks to bound $E_J$ and $E_T$.
• Applies parametric forms of the source error exponent via Blahut’s method to compute the lossy JSCC error exponent.
• Employs duality and convexity arguments to prove $E_T \leq E_J \leq 2E_T$ and conditions for equality and strict inequality.

Experimental results

Research questions

• RQ1 What is the exact value of the joint source-channel coding error exponent $E_J$ for a given discrete memoryless source and channel pair?
• RQ2 How does $E_J$ compare quantitatively with the tandem coding error exponent $E_T$ in terms of reliability and power efficiency?
• RQ3 Under what conditions does $E_J > E_T$, and when is $E_J = 2E_T$?
• RQ4 Can the JSCC error exponent be computed efficiently for arbitrary source-channel pairs?
• RQ5 What is the performance gain of JSCC over tandem coding in terms of power saving for practical systems like binary sources over noisy channels?

Key findings

• The joint source-channel coding error exponent $E_J$ satisfies $E_T \leq E_J \leq 2E_T$, establishing a theoretical upper bound on performance gain.
• For many source-channel pairs, $E_J$ is very close to $2E_T$, indicating substantial reliability improvement over tandem coding.
• Numerical results show a power saving of more than 2 dB for binary sources transmitted over additive white Gaussian noise and Rayleigh fading channels with finite output quantization.
• When the channel is symmetric, the bounds on $E_J$ admit closed-form parametric expressions, enabling efficient computation.
• The lossy JSCC error exponent under Hamming distortion is computed via a parametric form involving the source rate-distortion function and divergence minimization.
• Conditions for $E_J > E_T$ are derived based on the relative values of the source and channel error exponents, with strict inequality occurring when the source redundancy is effectively exploited.