Abstract: In particle filters (PF), sequential importance sampling will result in sample impoverishment and further the loss of diversity after resampling. Note that the signal points of modulation constellation are discrete and finite. The loss of diversity will result in performance degradation when PF is utilized to demodulate the multi-input multi-output (MIMO) systems. Therefore, an artificially noise enhancement (ANE) method is first introduced in importance sampling to increase the number of different particles, which will alleviate the sample impoverishment. Second, the partially deterministic reallocation (PDR) scheme is introduced to replace the resampling. The PDR scheme can increase the surviving probability of particles with small importance weights, thereby decreasing the loss of diversity. Analyses and simulation results show that the proposed method can significantly improve the bit error performance of MIMO systems.

Key words: multi-input multi-output, particle filters, sample impoverishment, loss of diversity