摘要 |
A detection unit obtains a position of a feature point of an occupant. A storage unit stores a model of all specific actions in each of physical feature classifications. The model defines a position or a locus of the feature point in each specific action, which an occupant may perform. A determining unit compares an in-action feature point with each model of a specific action, which is being performed, to determine a classification conformity weight indicating a possibility that the occupant belongs to the physical feature classification. An estimating unit compares the position of the feature point with each model to obtain a comparison result and generate a conformity value by incorporating the classification conformity weight into the comparison result. The estimating unit detects an estimated action, which is a specific action corresponding to the model, which has a highest conformity value. |