Abstract: A prototype insole sensor capable of measuring the gaits of both feet was developed. Feature analysis was performed using data acquired from simultaneous measurements of both feet. In addition, a dataset of peak values and rates of change was created from the acquired data and its application in machine learning was investigated. The results of the analysis using random forests showed that the rate of change was better at detecting detailed movements at the measurement points. The significance of the data obtained from the toes was confirmed. Therefore, we conducted an additional study using an insole sensor with more toe measurement points. Analysis of the rate of change showed the possibility of classification using toe-tip data. Therefore, random forest and principal component analyses were used. In this study, we demonstrated the usefulness of insoles for recording detailed individualized toe data. By accumulating and analyzing the data, we demonstrated the possibility of obtaining a guideline for the measurement location according to the subject's symptoms.

Keywords: Pressure sensors, Insole, Health condition change, Arduino, Bluetooth, Classification.

_________________________________________________________________________________________