Effective Cost Functions for Spectrum Entropy to Search for High-Frequency Event-Related Patterns in Electrograms with Noise
Abstract
Introduction: We discuss a wide range of problems about revealing hidden regularities in rearrangement of bio-electric activity of living organisms when it is registered on the background of various impacts, using look-up and temporal localization of event-related patterns in electrograms with noise. One of the approaches to solve such problems is based on Shannon entropy analysis calculated by the components of the power spectrum and called a function of spectrum entropy. When the sought patterns relate to high-frequency rhythms and their energy spectrum limits are a priori unknown, the cost functions of the spectrum entropy have low sensitivity. Purpose: Developing cost functions for entropy analysis which would have a sensitivity high enough to search for high-frequency patterns with a priori unknown parameters in electrograms with noise. Results: A cost function has been found which allows you to detect a frequency band corresponding to the maximum contribution of the spectral components of the sought patterns towards the total power of the spectrum. The subsequent calculation of the spectral entropy in the found frequency band provides a solution for the problem of finding event-related patterns under the conditions mentioned above. Practical relevance: The presented results confirm the effectiveness of using the developed functions. The only restriction is that an electrogram must be recorded on several electrodes.Published
2018-04-01
How to Cite
Shcherban’, I., Kirilenko, N., & Shcherban’, O. (2018). Effective Cost Functions for Spectrum Entropy to Search for High-Frequency Event-Related Patterns in Electrograms with Noise. Information and Control Systems, (2), 8-17. https://doi.org/10.15217/issn1684-8853.2018.2.8
Issue
Section
Information processing and control