Neural-net-based receiver structures for single- and multiamplitude bandlimited signals in CCI and ACI channels

This paper presents analysis and performance-evaluation results for several neural-network-based nondecision-feedback receiver structures, which improve the performance of bandlimited single- and multiamplitude signals transmitted over additive interference channels, such as cochannel interference (CCI) and adjacent channel interference (ACI). In particular, we propose, analyze, and evaluate a training algorithm for Nyquist-filtered single- and multiamplitude signals, based upon a novel nonuniform signal-sampling technique. We also introduce a novel nonlinear activation function for multiamplitude signals and evaluate its performance via computer simulation and in conjunction with various bandlimited signaling formats, detection techniques, and neural-network structures. Bit-error rate (BER) performance-evaluation results of the proposed neural-network receivers for coherent and noncoherent detection of Nyquist- and Butterworth-filtered single- and multiamplitude signals have shown performance improvements in the presence of CCI and ACI.