| Resumo: | Equalization is needed to remove inter-symbol-interference(ISI) caused by multi-path fading channels in wireless digital communication systems. Blind equalization is beneficial to mitigate the effects of limited channel bandwidth as it does not requires a training sequence to be sent with the transmitted signal. Both blind equalization algorithms, CMA and MMA, requires a step-size μ, to update the tap weights of the blind equalizer. It is important to select a suitable μ as it determines the convergence rate of the algorithm and the feasibility of convergence. Hence, a performance analysis for the optimal μ is done for modulation types BPSK, 8PSK, 4QAM, 16QAM and 64QAM. Further more, the optimal μ is used to test for the optimal modulation type for each CMA and MMA algorithm. Next, the performance of neural network equalizers, feed-forward neural networks and long-short term memory neural networks, are implemented to compare their performance against CMA and MMA algorithms.
|
|---|