Robust wavelet-aggregated signal

The modification of a method of wavelet-aggregated signal for multidimensional time series of monitoring systems is considered which takes into account the existence of strong outliers of the wavelet coefficients values. This property follows from ability of wavelet expansions to accumulate maximum of information in a rather small amount of coefficients as well from the nature of data, in which the outliers are not stipulated by errors of recording systems. The property of wavelet coefficients distribution to be non-Gaussian requires updating the usual schemes of the multi-dimensional analysis of canonical and principal components, which would allow make conclusions of the analysis of wavelet coefficients statistically more reliable. One of the probable approaches to solve this problem is presented with its applications to the analysis of geophysical data (monitoring time series and seismic catalogs) and comparing the results with those obtained by applying previous non-robust approach