Channel Prediction Based On Sinusoidal Modeling
Long range channel prediction is considered as one of the most important enabling technologies to future wireless communication systems. The prediction of Rayleigh fading channels is studied in the frame of sinusoidal modeling in this thesis.
A stochastic sinusoidal model to represent Rayleigh fading channel is proposed. The average of the conditional power spectrum of this model is shown to be the well known Jake's model. Given Doppler frequencies to be deterministic, the Cramer-Rao Lower Bound (CRLB) for the frequency estimates is derived. An algorithm to calculate the compressed CRLB is also proposed.
Using measurement data, the Jake's model is confirmed by the Normalized Mean Doppler Spectrum (NMDS) in both urban and suburban environments. The analysis of the time varying property of the model parameters shows that the model parameters are more consistent in suburban than in urban environment. A strong dominant sinusoid was observed in most suburban measurements, which might be due to the direct path in Line-Of-Sight (LOS).
Based on the statistical sinusoidal modeling, three different predictors are proposed. These methods outperform the standard LP in Monte Carlo simulations, but underperform with real measurement data. A subjective study of the LMMSE prediction methods to nearby tones are performed by simulations. The Unconditioned LMMSE predictor is found to be more suitable for the prediction of closely separated sinusoids.
Later, a Joint Moving Average and Sinusoidal (JMAS) model is proposed for channel prediction, which predicts the channel by LP and sinusoidal predictor jointly, together with a simple SVD based the ith biggest gradient model selection method. This method is termed Joint LMMSE predictor. It outperforms all the other predictors in suburban environments, but still performs slightly worse than the standard LP in urban environments.