A method of establishing a prediction model for each gait phase was adopted in order to achieve more accurate prediction of continuous movement of lower limbs for specific individual movement characteristics and use a shorter time to train the prediction model. The prediction model of the current phase was used to predict the joint angle after identifying the current gait phase. Support vector machine (SVM) was used to verify the proposed method. The experimental results show that the prediction method of continuous motion of lower limbs based on phase division has higher prediction accuracy and shorter model training time than the prediction method of joint angle modeling of the entire motion state. The correlation coefficients between the predicted results of the hip, knee, and ankle joints and the true values are all greater than 0.99. The average root mean square error of the predicted angle values and the true values is less than 2° each time, and the training time is shortened by 4.0~5.0 times.