ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 9. Vol. 26. 2020

DOI: 10.17587/it.26.507-514

O. G. Shcherban, Assistant Professor, e-mail: shchero@mail.ru, I. V. Shcherban, Professor, e-mail: shcheri@mail.ru, Southern Federal University, Rostov-on-Don, P. V. Lobzenko, Assistant Professor, e-mail: pasha.van@list.ru, Moscow Technical University of Communications and Informatics

The Real-Time Method for the Temporal Localization of the High-Frequency Patterns in Noised Multichannel Signals

The problems of the real-time searching and temporal localization of the short-time patterns in the multichannel signals are important in many practical cases. The desired patterns can be masked by the high-frequency noise and low-frequency artifacts. The known approaches of the high precision searching for the low-intensity patterns can be used during an a posteriori analysis of the signals. The real-time method for the searching and temporal localization for the short-time low intensity patterns in multidimensional dynamic objects noised output signals has been considered. We synthesized an adaptive multichannel singular spectrum analysis (MSSA) band-pass filter (BPF). Based on the power spectral density of the signal amplitude the grouping rule determines adaptively the MSSA reconstructed components as low-frequency artifacts or high-frequency noise and removes them. The grouping rule thus enables BPF to be adaptive to the output signals containing different levels of the low-frequency artifacts or high-frequency noise. This provides the required sensitivity of the criterial function for the real-time searching and temporal localization for the low intensity patterns. The distance Hausdorff metric was used to measure the similarity of the two current neighboring non-overlapping time samples from the multichannel output signals. A computational experiment results illustrating the efficiency of the use of the method developed has been given. The eight channel electroencephalography (EEG) signal based on the Markov Process Amplitude EEG model were used to verify the developed method. Results showed a better performance in the real-time searching of the low intensity patterns, compared with the method, based on the Butterworth bandpass Fourier filter.
Keywords: real-time searching, short-time patterns, output signals, multidimensional dynamic objects, multichannel singular spectrum analysis, adaptive bandpass filter, the Hausdorff metric

To the contents