Theory of Local parameters in Time-Frequency Analysis and its Applications to Tracking States of Mental Stress

点击翻译按钮获取中文摘要

The state of mental stress is often measured and estimated by using EEG and/or ECG. Since feature of these signals and the mental state change with time, we have to deal them as non-stationary data. Therefore we have to develop effective methods for non-stationary data analysis.The theory of time-frequency analysis is known as a powerful tool to analyze non-stationary signals like biomedical signals.In this project, we tried to find effective parameters by which we can extract essential characteristics of a random process and analyze and interpret the non-stationary signals. In the theory of random processes, many kind of effective bandwidths have been proposed and used very often in may branches. How ever there have been few effort to unity them. We proposed a general class of infinite effective bandwidths which includes all the essential traditional bandwidths. The class is defined by Renyi's alpha-order entropy and gives unified expression of the bandwidths. We call them alpha-order bandwidths, W-alpha. The concept of W-alpha is easily expand to non-stationary analysis. Then we proposed time varying W-alpha based of time-frequency distributions and showed that the time varying W-alpha is effective by simulation analysis. We applied them to EEG analysis measured from persons who are put mental stress by been a task of Uchida-Kraepelin. We showed the method is useful to track alpha-bandwidth of EEG.Meteorological conditions are essential factors which affect human mental state. In the late stage of this project, we began to develop methods to analyze meteorological data, that is, image data from some meteorological satellites, temperature data, rainfall data, etc. We tried apply information theoretical concepts, like entropy and divergence to our methods. We have been showing these information theoretical concepts are very powerful to our purpose.