Coherent Detection of Signals with Time-Varying Parameters Using Multiple Observations

This paper investigates the multiple observation detection statistic computed as the average of a non-overlapping time-sequence of magnitude-squared coherence (MSC) statistics, denoted AMSC. The multiple observation statistic provides a means to counteract the decrease in MSC detection performance caused by variation of the unknown signal parameters. This variation may be attributed to boundary changes and propagation channel nonstationarities over the integration interval. Performance of the MSC detector is compared with the AMSC detector for equivalent integration times using receiver operating characteristic (ROC) curves.