The Multi-Channel Maximum-Likelihood (MCML) method for infrasound detection and wave parameter estimation: Evaluation, implementation, and first applications

Abstract

The detection capability of the International Monitoring System (IMS) infrasound network is a key concern since ubiquitous coherent noise interfere with the identification of explosive events. The need to distinguish between incoherent wind noise and real coherent infrasonic wave signals has motivated the development of a new parametric stochastic model. We establish the expressions of both the Generalized Likelihood Ratio Test (GLRT) and the Maximum Likelihood Estimator (MLE). The major results are the expression of the asymptotic distribution of the GLRT under null hypothesis, leading to the p-value computation, and the expression of the asymptotic covariance. The Multi-Channel Maximum-Likelihood (MCML) detection is implemented in the time-frequency domain in order to discriminate between interfering signals. Extensive simulations with synthetic signals show that MCML outperforms the state-of-the-art multi-channel correlation detector algorithms in poor signal-to-noise ratio scenarios. We illustrate the performance of MCML on real IMS data, highlighting its capability to characterize the coherent ocean ambient infrasound noise in high resolution through comparisons with theoretical state-of-the-art numerical wave models as implemented at the LOPS research unit of IFREMER in the DATARMOR HPC center.