Detection of short transients in colored noise by multirate multiresolution (MRMR) analysis

Based on a multirate multiresolution (MRMR) Kalman filter analysis, two de-noising methods are applied to a colored noise model to detect short transients. Starting with a standard state-space stochastic model, it can be shown that a signal can be decomposed into uncorrelated innovations at varying sampling frequencies and corresponding levels of resolution. In a benchmark experiment, results show that as a signal is decomposed in a wavelet packet, transient signals can be identified at lower thresholds. For applications in undersea warfare, this detection scheme can provide valuable information in a sensor fusion design.