Non-Gaussian signal detection from multiple sensors using the bootstrap

Existing tests based on the cross bispectrum to detect stationary non-Gaussian signals use two sensors or channels of data. We propose to extend such tests to the case of multiple sensors. Our approach uses Bonferroni tests of multiple hypotheses. A multi-sensor bootstrap method is presented and compared through simulations with two other multi-sensor methods. Simulation results show that the bootstrap method is better able to keep the level of significance and have high correct detection (as the SNR increases) than the others.