To recognize human behavior, foot pressure sensing shoes were developed, and an analysis for discriminating the user's actions from the foot pressure distribution was examined, considering the movements, walking, running, standing, sitting, going upstairs and downstairs, and cycling. These seven actions were discriminated with a high accuracy using feature quantities such as the average, standard deviation, maximum, and difference deviation extracted from the data of only three sensors by discriminant analysis. In addition, by canonical discriminant analysis, discriminant functions which classify the seven actions with an accuracy of 100% were derived by using feature quantities extracted from the data of five sensors or less.
View full abstract