The method of reducing the noise influence and time-series samples fluctuations on the accuracy of information representation

The paper considers a method of reducing the error in the representation of information in telecommunication monitoring systems by filtering noise and fluctuations in the levels of samples of time series. To solve this problem, the wavelet-processing method is used, in particular, the effect of time-series fluctuations is reduced by calculating approximating coefficients of the n-th order, which corresponds to performing multilevel statistical processing of the values of the time series readings and is equivalent to passing the signal through the low-pass filter. In addition, the accuracy of information representation is enhanced by suppressing the components of the noise process by performing a tresholding operation on the detailing coefficients up to the n-th level of the decomposition. In this case, it is proposed to implement the trsholding algorithm according to the criterion of the adaptive penalty threshold.