To provide a passenger state detection method capable of automatically generating a recognition argorithm and highly accurately estimating the state of a passenger sitting on a seat, and a passenger state detection apparatus.

The passenger state detection method comprises the steps of measuring (S1) a pressure pattern by a plurality of sensor sections provided on a seat section where a pressure is induced when the passenger takes a seat, inputting the pressure pattern measured as above to an SVM for executing (S2) extraction of features by the SVM, and discriminating by an SVM separation curvature face to discriminate (S3) classes as to adult/child of the passenger. Therefore, the state of the passenger sitting can be estimated with a high accuracy non-invasively by using the pressure pattern data. Because the recognition algorithm is automatically generated by learning, man-hour for designing can be reduced.