In human sitting position recognition research based on resistive thin film pressure sensor

the arrangement of pressure sensors is a key problem

which affects the recognition accuracy and hardware cost. A sensor arrangement and selection algorithm is proposed based on Euclidean distance similarity measurement of time series. The normalized pressure time series data of different postures is defined. In the cross-contrast experiments

the random forest and stack support vector machine are utilized. Expenmental results show that 25 percent sensors are reduced but the accuracy of recognition is improved. The effectiveness of the algorithm is validated.