Semi-blind approach to second order identification of SIMO-FIR channel driven by finite-alphabet sequence

This contribution is concerned with the problem of joint single input multiple output (SIMO)-finite impulse response (FIR) channel/sequence estimation when the transmitted sequence contains both information and training parts. The ordinary technique in such a case is training part-based MMSE estimation of the channel IR with subsequent recovery of the information part via a Viterbi algorithm. We propose an approach to improve the performance of channel identification by exploiting the information part of the transmitted sequence for channel IR estimation. This approach can also be seen as an improvement of a blind technique through the use of a training sequence. Computer simulations show better performances of the new techniques compared to the related MMSE and purely blind conditional maximum likelihood-based approaches such as iterative least squares with projection (Talwar et al., 1995)