Improving precision of three-dimensional ground reaction force(GRF)estimation based on local plantar pressure can assist in clinical diagnosis of foot diseases and evaluation of rehabilitation.An enhanced method is proposed:estimating model for three-dimensional ground reaction is constructed.Plantar local pressure measured by wearable collection device is Gaussian derived by Legendre pseudo-spectral interpolation method to simulate distribution of plantar pressure of relevant zone

model for three-dimensional ground reaction force estimation is constructed and fused with triaxial acceleration data of lower limbs through artificial neural networks.The normalized root mean square error(NRMSE)for vertical(V)

anterior-posterior(A/P)and medial-lateral(M/L)ground reaction force estimations are 7.48%

7.11% and 8.62% respectively

which have higher estimation precision compared to the methods that solely relying on local plantar pressure.