Analyzing Video Information by Monitoring Bioelectric Signals

摘要

The paper considers the problems of creating tools for video information flow automatic analysis through monitoring certain characteristics of bioelectrical signals of an agent (expert, operator). The authors used spectral analysis methods to obtain time series that illustrate power changes of experimental bioelectric signals (Sa) of an agent during studying a continuous video stream. The fuzzification of spectral features allows moving to fuzzy time series illustrating information evaluation by an agent. The spectra are calculated using a sliding working window from fragments of two types of bioelectric signals, which are recorded synchronously in two autonomous channels. Based on fuzzy evaluations of spectral features and Mamdani fuzzy inference algorithm, the algorithm for analyzing video information allows classifying video fragments according to the sign of agent's emotional reaction. Tsukamoto algorithm localizes time markers that determine the beginning and the end of each fragment.