A Monte-Carlo Implementation of the SAGE Algorithm for Joint Soft Multiuser and Channel Parameter Estimation

TL;DR

This paper introduces a Monte Carlo-based SAGE algorithm for joint estimation of transmission delays and channel parameters in uplink DS-CDMA systems, utilizing Gibbs sampling for efficient computation.

Contribution

It presents a novel MCMC-SAGE algorithm that combines Monte Carlo methods with the SAGE framework for improved joint parameter estimation in asynchronous DS-CDMA.

Findings

Excellent estimation performance demonstrated in simulations

Guaranteed convergence in likelihood of the proposed algorithm

Efficient computation via Gibbs sampling

Abstract

An efficient, joint transmission delay and channel parameter estimation algorithm is proposed for uplink asynchronous direct-sequence code-division multiple access (DS-CDMA) systems based on the space-alternating generalized expectation maximization (SAGE) framework. The marginal likelihood of the unknown parameters, averaged over the data sequence, as well as the expectation and maximization steps of the SAGE algorithm are derived analytically. To implement the proposed algorithm, a Markov Chain Monte Carlo (MCMC) technique, called Gibbs sampling, is employed to compute the {\em a posteriori} probabilities of data symbols in a computationally efficient way. Computer simulations show that the proposed algorithm has excellent estimation performance. This so-called MCMC-SAGE receiver is guaranteed to converge in likelihood.