Time Delay Estimation for Noise-Like Signals Embedded in Non-Gaussian Noise Using Pre-filtering in Channels

Abstract

A task of time delay estimation for two sensors that receive a noise-like wideband signal embedded in non-Gaussian environment is studied. One peculiarity of time delay estimates for conventional method of cross-correlation data processing is that they can be abnormal due to intensive noise influence and a limited interval of signal observation. Performance can be improved by pre-filtering of signal/noise mixtures in both channels. In addition to center weighted median filter proposed earlier, we consider application of 1-D filter based on discrete cosine transform (DCT). It is shown that the use of such a combination of center-weighted median and DCT-based filters provides more robust estimation in conditions of low signal-tonoise ratios (SNRs).