TIME SERIES PROCESSING USING WAVELET-TRANSFORMATIONS FOR ACCURACY INCREASE IN INFORMATION PRESENTATION
The purpose of this work is development of the method for error decrease in information presentation in telecommunication systems of monitoring by means of filtering noise and fluctuations of levels in time series counts. To solve this problem there is used a method of wavelet processing. In particular, the decrease of time series fluctuation impact is carried out by means of the computation of approximating coefficients of the n-th level which corresponds to the fulfillment of multi-level statistical processing the values of time series counts and equivalent to a signal passage through a filter of low frequencies. There was developed and investigated a simulator and its statistical parameters of processing with a wavelet transformation of time series counts. It is shown that time series wavelet processing and the application of approximation coefficients of waveletdecomposition increase the accuracy of data presentation. It is also ensured at the expense of noise component suppression through a method of thresholding upon detailing coefficients of decomposition. In the paper there are shown investigations of the dependence of approximation coefficient correlation time upon a wavelet decomposition level. There was also investigated a depression dependence of noise components of time series count fluctuations of emission at the processing with the wavelet decomposition with obtaining approximation coefficients of different levels. The fulfilled analysis of the results of different criteria application and approaches to smoothing on the basis of threshold processing the detail coefficients of wavelet decomposition has shown that at smoothing time series there will be an optimum choice of an adaptive penalty threshold level. The presented results of smoothing with an adaptive penalty threshold have shown that the signal-noise ratio increased for more than 2.53dB in comparison with the initial one.