Comparison of linear and neural network-based power prediction schemes for mobile DS/CDMA systems

This paper presents a novel neural network based predictor for received power level prediction in direct sequence code division multiple access (DS/CDMA) systems. The predictor consists of a functional link neural network (FLNN) followed by a multilayer perceptron (MLP). An important but difficult problem in designing the cascade predictor is to determine the complexity of the networks. We solve this problem by using the predictive minimum description length (PMDL) principle to select the optimal number of input and hidden nodes. This results in a predictor with both good noise attenuation and excellent generalization capability. The optimized neural network predictor is compared with a class of FIR predictors with simulations employing noisy Rayleigh fading signals with a 1.8 GHz carrier frequency. The results show that the neural predictor can provide a smoothed output signal with a signal-to-noise ratio (SNR) gain of about 10 dB, which outperforms the linear counterpart. This makes the neural predictor well suitable for applications where 'delayless' noise attenuation and efficient reduction of fast fading are required.