Interference Rejection for Spread Spectrum Signals Using the EM Algorithm

Spread-spectrum systems are well known to be robust to narrowband interference. Even for these systems, however, performance degrades substantially when interference levels rise, requiring the use of further signal processing to combat interference. Traditionally, the approach has been to excise the interference by either estimating it and then subtracting it, or by filtering it out. In this paper we instead formulate the problem as one of maximum-likelihood (ML) sequence estimation in the presence of interference, and then solve it by using the expectation-maximization (EM) algorithm. We look first at the problem of signle-tone interference, and then generalize it to the case of narrowband Gaussian interference. Comparisons are made to the optimum receiver, and to a receiver that utilize a notch-filter to reject the interference. The EM algorithm is seen to perform esentially optimally, even for large interference levels.