This thesis investigates the problem of adaptive equalization in the presence of Inter-Symbol-interference (ISI), additive white Gaussian noise, cochannel and adjacent channel interference. The thesis proposes a new equalizer based on the neural multi-layer perceptron structure. The proposed neural equalizer is termed: Multi-Layer Perceptron Decision Feedback Equalizer (MLPDFE). It is designed to realized an adaptive equalizer with non-linear decision boundaries. The new equalizer is capable of operating under poor Signal to Interference Ratio (SIR) and poor Signal to Noise Ratio (SNR).

This thesis starts by demonstration of the sensitivity of the MLPDFE performance (in terms of average Mean Square Error MSE) on its perceptron configuration and adaptive parameters. The effect of co-chanel and adjacent channel interference on the system performance is investigated separately.

The thesis presents the results of a computer simulation designed to compare the performance of the new MLPDFE with that of a conventional Least Mean Square (LMS) Decision Feedback Equalizer (LMSDFE) in terms of the error performance. The results indicate that the MLPDFE provides better symbol error rate relative to the LMSDFE when only the cochannel interference is dominant. In cases when the cochannel interference is less than the combined effects of all other impairments, the performances of MLPDFE and LMSDFE are very similar.