Second-order asynchronous interference cancellation: Regularized semi-blind technique and non-asymptotic maximum likelihood benchmark

An asynchronous interference cancellation problem is addressed where training and working intervals are available that contain the desired signal and arbitrary overlapping interference. A regularized non-iterative estimation of the antenna array coefficients is proposed which employs the autocorrelation matrix estimation as a weighted sum of the autocorrelation matrices estimated over the training and working intervals. A non-asymptotic second-order statistic maximum likelihood (ML) benchmark is developed, which is based on estimation of the structured covariance matrices over the training and working intervals for the Gaussian data model. It is shown by means of simulation in the TDMA and OFDM environments that the regularized semi-blind solution significantly outperforms the conventional estimators and demonstrates performance close to the proposed benchmark.