基于鞋垫传感器测量垂直地面反作用力的算法

目的 提出一种实用、高精度的足底压力动态监测算法，通过电容式鞋垫传感器测量走路时垂直地面反作用力(vertical ground reaction force, vGRF),并验证预测精度的可靠性。方法 4名健康男性受试者穿戴电容式鞋垫传感器，在Kistler三维测力台上采集快速和慢速行走数据。对电容式鞋垫传感器采集到的数据进行像素化处理，处理后的数据输入残差神经网络ResNet18,预测得到高精度vGRF。结果 与Kister测力台收集的数据分析比较，快速和慢速行走的归一化均方根误差(normalized root mean square error, NRMSE)分别为8.40%和6.54%,皮尔森相关系数(Pearson correlation coefficient)均大于0.96。结论 本研究为移动场景下GRF动态测量提供了一种新型算法，可以用于实验室外完整GRF估计，而不受测力台数量和位置的约束，潜在应用领域包括步态分析以及有效捕捉病理性步态。

An Algorithm for Measuring Vertical Ground Reaction Force Based on Insole Sensors

Objective A practical and highly accurate algorithm for dynamic monitoring of plantar pressure was proposed, the magnitude of vertical ground reaction force (vGRF) during walking was measured by a capacitive insole sensor, and reliability of the prediction accuracy was verified. Methods Four healthy male subjects were require to wear capacitive insole sensors, and their fast walking and slow walking data were collected by Kistler three-dimensional (3D) force platform. The data collected by the capacitive insole sensors were pixelated, and then the processed data were fed into a residual neural network, ResNet18, to obtain high-precision vGRF. Results Compared with analysis of the data collected from Kister force platform, the normalized root mean square error (NRMSE) for fast walking and slow walking were 8.40% and 6.54%, respectively, and the Pearman correlation coefficient was larger than 0.96. Conclusions This study provides a novel algorithm for dynamic measurement of GRF in mobile scenarios, which can be used for estimation of complete GRF outside the laboratory without being constrained by the number and location of force plates. Potential application areas include gait analysis and efficient capture of pathological gaits.