Estimating GRF(Ground Reaction Force) and Calibrating CoP(Center of Pressure) of an insole measured by an low-cost sensor with neural network

CoP(Center of pressure) and GRF(ground reaction force) of insole are very important values in biomechanics area. They are using for calculating kinematics, dynamics of human or controlling of robot like exoskeletons. As an alternative to high-cost insole pressure sensors that can measure the insole pressure distribution and calculate the center of pressure, a FSR (Force Sensing Resistor) foot sensor with FSR sensors on the bottom of the insole was developed. However, the value of the CoP calculated using fixed coordinates and the values of FSR sensors were not sufficiently accurate and FSR sensors cannot cover the whole area of the insole so it can not calculate the magnitude of GRF. Hence, in this paper, a model capable of estimating of GRF and calibrating CoP measured by FSR foot sensors using neural network fitting is introduced. These processes rely on the fact that foot has protruding areas that are initially in contact with the ground while walking, with the size and magnitude of the pressure exerted by other non-protruding areas estimated using the the constant patterns of the pressure values of the protruding areas. This paper presents the division of the insole based on anatomical shape of foot, estimations of appropriate numvers and locations of the FSR sensors, creation of virtual forces and their floating coordinates, development of algorithms with neural network fitting for estimating the values, and calculation of the estimated GRF and calibrated CoP. Validation is conducted by comparing the Values with those of F-Scan System(Tekscan, Inc.).