Channel prediction for link adaptation in fast fading environments

J.F. SCHMIDT; J. E. COUSSEAU; R.WICHMAN

Rio Gallegos

Congreso; XII Reunión de Procesamiento de la Información y Control; 2007

RPIC

We present a channel predictor basedon a basis function approach formulation of a Kalmanfilter channel estimator. The estimator models channelvariations via truncated discrete cosine transform(DCT), which is robust to the shape of the channelDoppler spectrum. We show that the DCT can be approximatedby a set of linear narrow passband filtersthat can be formulated as Kalman filters that tracka set of decoupled parameters. Since the variationof these parameters is much slower than the channelfading, and the modeling accuracy provided by the filterbank is much better than with traditional models,good prediction performance is achieved by extrapolationin time. Since Doppler frequencies of interestare much lower than the signal bandwidth, a broaderprediction horizon is obtained by decimating the estimatedchannel samples and scaling up in frequencythe filter bank accordingly. The derived predictor attainsless than 10% power prediction error with a predictionhorizon of 1 Doppler wavelength (more thanone data frame prediction range).