Automation of greedy algorithm for segmentation with minimized exception : binary for EEG burst detection homogenous string segmentation

Abstract

A survey has been carried out to investigate, various homogeneous criteria and algorithms used for string segmentation. An efficient implementation of greedy algorithm for SME – Binary (segmentation with minimized exception) is carried out for EEG signal burst detection. The result of greedy algorithm is analyzed in the context of successful burst detection for various values of threshold and number of segments. An automation of this greedy algorithm is purposed for the burst detection in EEG signal. The results obtained form this automation is analyzed. Greedy algorithm for SME-Binary is able to detect bursts. But it is not very successful for a single value of threshold. It is also not known for what value of k (number of segments in a given string), the algorithm will successfully recognize all bursts as k is an input parameter to this algorithm. Automation of this greedy algorithm is quite successful in recognizing the bursts in the EEG signal. The main advantage is, it is independent of value of k and threshold. This algorithm is also quite fast in detecting bursts. The disadvantage is, some bursts are recognized by number of segments rather than a single one. So to count the number of bursts a clustering algorithm is required to enhance this automation. This is purposed as a future work in this report.