Detecting and identifying anomalous effects in complex signals

We propose a method for detecting and identifying anomalous effects in a signal of a complex structure based on nonlinear approximating schemes in the dictionary of wavelet packets. Taking into account the properties of the time–frequency window of the wavelet transform, an adaptive threshold is introduced. Increasing the efficiency of detecting various types of structures is achieved by applying a superposition of wavelet transform constructions. Using neutron monitor data as an example, it is shown that the method permits one to suppress noise and identify anomalous effects of various shapes and duration. The results confirmed the efficiency of the proposed method for detecting low-amplitude Forbush effects in cosmic ray variations.