Multiuser detection based on hidden Markov models

For investigation on the slow convergence speed of channel estimation and the unreason of state estimation in the probabilistic algorithms, a new multiuser detection was presented. Based on the hidden Markov models, it adopted the fixed-lag to avoid divergence in the estimation channel, and used the sawtooth-lag scheme in calculating the probability of the backward sequence to decrease computational complexity. With the correlation between the previous and the following sequence and the steepest descent algorithm, the new algorithm utilized DLMS to estimate the mutlichannel of the multiuser. Considering seeking for the following sequence of maximal transition probability corresponding to previous sequence, a maximum a posteriori (MAP) multiuser detector had been constructed. The computer simulation illustrates that the new algorithm increase the speed of channel estimation, global convergence and system stability.