H2 optimal signal predictive quantization in oversampled FBs

This paper is concerned with the design of signal predictor in oversampled signal predictive subband coding. Spectral analysis of the prediction error system is used to derive an H2 optimization formulation of the design problem, and an LMI based method is obtained to design the H2 optimal signal predictor that minimizes the variance of quantizer input signals. Simulation examples are presented to demonstrate the obtained design method.