Using an autoregressive model for DMC

In this work an iterative two-step procedure is proposed for estimating the diffuse multipath component (DMC) in noise and specular multipath components from channel measurement data. During the first step of the scheme the specular part of the impulse response from the measured data is estimated. Then, the reconstructed channel impulse response due to specular components is subtracted from the measured channel data to calculate the residual. It is assumed that the obtained residual signal contains DMC in white measurement noise. By modeling the DMC as colored noise process, the response of the channel due to DMC is then estimated using a multichannel autoregressive (MAR) model as a second step. The two steps, estimating the specular components and the DMC is then repeated until a combined convergence criterion is reached. By validating the proposed estimation scheme using measured channel data, it was observed that the final residual including the MAR model is white. This indicates that the correlations in the residual are significantly captured by the MAR model.