Search for gravitational wave bursts by wavelet packet decomposition: The detection algorithm

We present a novel method based on wavelet packet transformation for detection of gravitational wave (gw) bursts embedded in additive Gaussian noise. The method exploits a wavelet packet decomposition of observed data and performs detection of bursts at multiple time-frequency resolutions by the extreme value statistics. We discuss the performances of detection algorithms (efficiency and robustness) in the general framework of hypothesis testing. In particular, we compare the performances of wavelet packet (WP), matched filter (MF), and power filter (PF) algorithms by means of a complete Monte Carlo simulation of the output of a gw detector, with the detection efficiencies of MF and PF playing the role of upper and lower bounds, respectively. Moreover, the performances of impulsive filter (IF) algorithm, widely used in the data analysis of resonant gw detectors, have been investigated. Results we get by injecting chirplet signals confirm the expected performances in terms of efficiency and robustness. To illustrate the application of the new method to real data, we analyzed a few data sets of the resonant gw detector AURIGA.