Human's operation modeling based on Hidden Markov Model with consideration for the operation change and timing

Abstract

A new human's operating model using Hidden Markov Model (HMM) is proposed to predict man-machine communication. In the proposed method, human's next operation is predicted by the current human operation and the timeline data of the environmental change. The prediction of operation requires two functions: the correctness of predicted change and timing. In conventional study, since the HMM consists of a dataset which data length are not constant, the predicted operation timing is inarticulate even if the estimated change is correct. To solve this problem, we propose to divide the data by a constant length. In this method, the self returning operation change is also modeled and the operation change timing is guaranteed within the divided data length. The data length is defined for the operational characteristic by preliminary check of human operational data. In this paper, the length is set to the frequently-appearing minimum length. By using the data handling and modeling, the operating model can predict precisely the change and the timing. The soccer game is made on PC to evaluate if the proposed method leads to the precise prediction. Finally, we compared the conventional method and the proposed method.