A Novel Adaptive Multilevel - Quadrature Amplitude Modulation (M-QAM) Receiver Using Machine Learning to Mitigate Multipath Fading Channel Effects

The demand for faster speed and greater signal strength of today’s wireless broadband technology evolved signaling techniques to improve spectral bandwidth efficiency. In wireless digital communications, higher-order of M-QAM technique has been employed to improve bandwidth and channel efficiency of the signaling. However, wireless broadband communications such as mobile phones and wireless access networks are prone to multipath fading channel. Higher-order M-QAM is very susceptible to this fading channel because it affects both the amplitude and carrier phase of the transmitted signal as it induces non-linearity. In this study, an innovative approach in M-QAM demodulation technique has been proposed. This thesis investigated the performance of a proposed modified Costas Loop M-QAM receiver that employed Machine Learning using multi-layer perceptron (MLP) with error-back propagation (EBP) as an adaptive amplitude fading estimator and, using fuzzy logic as loop filter in the phase lock loop (PLL) circuit that estimated the distorted carrier phase of a received signal. A computer simulation of the proposed receiver was developed to investigate the performance of the receiver’s signal recovery over Rayleigh fading channel. The results showed that the ML algorithm tracked well the phase noise and the bit-error rate (BER) were comparable to the theoretical M-QAM curve on certain values of signal-to-noise (SNR) levels