Robotic knee prosthesis plays an important role to assist amputation subjects while recovering in daily living activitiesand human lower limb motion intention mapping Based on surface electromyography(sEMG) is the key step towards the controlof robotic knee prosthesis.In this paper, we propose a method for predicting knee joint angles base onsEMG.Predict the knee joint angle by extraction of the time domain features of sEMG and building the mapping relationship between sEMGand knee joint angle during flat walking through BP neural network model.