Comparison of filters for detecting gravitational wave bursts in interferometric detectors

Abstract

Filters developed in order to detect short bursts of gravitational waves in interferometric detector outputs are compared according to three main points. Conventional Receiver Operating Characteristics (ROC) are first built for all the considered filters and for three typical burst signals. Optimized ROC are shown for a simple pulse signal in order to estimate the best detection efficiency of the filters in the ideal case, while realistic ones obtained with filters working with several "templates" show how detection efficiencies can be degraded in a practical implementation. Secondly, estimations of biases and statistical errors on the reconstruction of the time of arrival of pulse-like signals are then given for each filter. Such results are crucial for future coincidence studies between Gravitational Wave detectors but also with neutrino or optical detectors. As most of the filters require a pre-whitening of the detector noise, the sensitivity to a non perfect noise whitening procedure is finally analysed. For this purpose lines of various frequencies and amplitudes are added to a Gaussian white noise and the outputs of the filters are studied in order to monitor the excess of false alarms induced by the lines. The comparison of the performances of the different filters finally show that they are complementary rather than competitive.PACS numbers 04.80.Nn, 07.05.Kf

The gravitational wave 'probability event horizon' for double neutron star mergers

“…We assume optimal (matched) filtering can be applied to the candidate transients, but the filters that will be most effective will not be exactly matched to individual events. The 'norm' and 'mean' filters, although not optimal, should perform at a significant fraction of a matched filter for transients where the signal is not completely characterized (Arnaud et al 2003). We emphasize that using the PEH in a signal processing context is based on fitting to outliers using a model distribution and not 'detecting' individual events.…”

“…Noise transients of varying amplitudes mimic sources at varying redshifts, enabling the introduction of a 'noise PEH'. Here we use a 'Gaussian-noise PEH'; this is based on a standard zero-mean Gaussian distribution with standard deviation s d = h rms √ f s /2 f c , modelled on idealized interferometric detector noise with h rms denoting the root-mean-square amplitude of the noise and f s the sampling frequency; see Arnaud et al (2003). Fig.…”

Does transparent hidden matter generate optical scintillation?

“…For example, robust filtering developed in VIRGO for burst searches (Arnaud et al 2003) should be appropriate to the search for an oscillating signal within a definite short period. The time scales of the data flow and of the signal are just multiplied by five orders of magnitude with respect to VIRGO.…”

Benefits of joint LIGO - Virgo coincidence searches for burst and inspiral signals

“…The arrival time of the signal is defined as the time of the maximal amplitude. For the 1ms Gaussian signals, analyzed with the Peak Correlator, the arrival time can be determined to within 0.3 ms on average (the time accuracy obtained by the Peak Correlator on Gaussian signals can be parametrized as: σ P C = 1.4310 −1 10 SNR ms [12]). Taking into account the observed arrival times and their corresponding timing accuracy at each of the sites (the accuracies are computed using the detected SNR and the previous parametrization), we use a χ 2 minimization technique to determine the sky location [13].…”