Roan M.

It is common in Acoustics to measure a signal that has been degraded by propagation through an unknown, noisy channel prior to measurement. While only the degraded measured signal is available, the data of interest are the original signal and the channel parameters. Often, it is desirable to reverse the filtering process by application of an inverse filter to recover the original signal. When neither the input signal properties nor the channel properties are known, this process is known as blind deconvolution (BDC). Typically, BDC algorithms assume noiseless, stationary propagation channels and input sources. These assumptions are usually violated in practical applications (e.g., noisy multipath propagation environments with moving source and receiver). To model these effects, predictive techniques are applied to incorporate a priori information about the system into the existing blind processing framework. The original contributions of this work follow. First, a novel formulation of the extended Kalman filter (EKF) is proposed. This allows incorporation of a priori information into gradient-based blind processing algorithms. This formulation is then applied to the Natural Gradient (NG) BDC algorithm. Finally, results are presented that suggest significant improvement in signal recovery performance over the NG BDC algorithm for dynamic noisy channels.