Analysis of environmental electromagnetic signal using nonnegative Matrix Factorization minimizing quasi-L1 norm

Anomalous environmental electromagnetic (EM) radiation waves have been reported as the portents of earthquakes. We have been measuring the Extremely Low Frequency (ELF) range all over Japan. Our goal is to predict earthquakes using EM radiation waves. The recorded data often contain signals unrelated to earthquakes. These signals, as noise, confound earthquake prediction efforts. It is necessary to eliminate noises from observed signals in a preprocessing step. In previous researches, we used ISRA, an algorithm of the Nonnegative Matrix Factorization (NMF), to estimate source signal. However, ISRA is not robust for outliers because ISRA's cost function is based on square distance. In order to improve robustness, we should use lower order cost function. In this paper, we propose a nonnegative matrix factorization method using quasi-L1 norm in cost function (quasi-L1 NMF). In the experiment using ELF observed signals that include outliers, the proposed method extracts source signals more accurately than ISRA.