An off-line self-calibration method for resistive insole pressure sensors

Accurate and lightweight acquisition of human plantar pressure information such as vertical ground reaction force (vGRF) is crucial in applications such as clinical analysis and human-machine collaborative control. In recent years, resistive thin-film insole force sensors have become a representative solution, with advantages such as low cost, portability, and easy development. As the sensor is used for longer periods of time, more significant data drift is a problem that this insole sensor still needs to solve. In the subject experiment, the average root mean square error (RMSE) of the insole sensor measurements during the stance phase of the gait cycle which relative to the treadmill force platform measurements is approximately 44.7% of body weight, and the bimodal characteristics of vGRF in the stance phase were difficult to identify. To address this problem, this study demonstrates a low-cost offline self-calibration method for insole force sensors based on optimization, aiming to help users easily and efficiently improve the data accuracy of insole force sensors when walking on flat ground outdoors. The study built a wearable sensor system for experimental data collection and calibration method verification to autonomously collect the data required for the experiment. After the self-calibration method was implemented, the RMSE of the vGRF output by the insole force sensor during the support phase was approximately 10.2% (±0.6%) of the body weight, which was an average reduction of 77.2% in the error before calibration. The bimodal features corresponding to the vGRF could be significantly identified during walking. The results of this study show that this offline self-calibration method for insole force sensors can significantly improve the accuracy of plantar pressure information.