Joint Channel and Symbol Timing Estimation and Data Detection

We consider the problem of data detection in the absence of channel state, symbol timing and noise power information. We propose a system parameter estimation method based on a two-step implementation of the Expectation-Maximization (EM) algorithm. Due to the sampling time offset, the received sequence is correlated. As a result the EM algorithm cannot be directly implemented. We divide the received sequence into two (odd and even) subsequences so that the symbols in each subsequence are independent. The EM algorithm is then devised for each subsequence and is applied iteratively, switching from one subsequence to the other after each iteration. The algorithm estimates the channel coefficient, the noise power, and the sampling time offset from each frame of data. The estimated parameters are used in the (maximum a posteriori) detection of the data sequence using the Viterbi algorithm. Numerical results from the proposed algorithm show a very good match with those for a system that has knowledge of all the system parameters.