Data-Aided Channel Estimation for Poisson Channels With Inter-Symbol Interference

It has been shown that long pilots should be employed in a Poisson channel with inter-symbol interference to obtain accurate channel estimate, which may reduce the transmission efficiency. In this paper, an iterative data-aided channel estimation and signal detection algorithm is proposed with significant shortened overhead but can approach the bit-error rate performance under perfect channel state information. In this algorithm, periodic pilots are not required and only one short pilot sequence is utilized to obtain a rough initial channel estimate, which will be strengthened and updated by iteratively employing the blockwise data symbols as additional pilots. It is proved that treating the detected data symbols as pilots is a biased estimation, where the bias is proportional to the bit-error rate. Numerical results show that the proposed approach employing only 20 initial pilots can approach the optimal bound. The same bound needs to be achieved by assigning at least 500 pilots using periodic pilot based channel estimation.