Abstarct: Rehabilitation is one of the extensive subsets of robotics, and physiotherapy is one of the most common methods of rehabilitation. Muscle force estimation has many applications in physiotherapy, rehabilitation and making auxiliary equipment. Estimated force using surface electromyography (sEMG) signals used, in some physiotherapist assistant robots, to control the robot, diagnose the disease, determine the type of treatment and physiotherapy methods. In this study, a model baxsed on support vector regression (SVR) is proposed in two ways and to estimate muscle force with surface electromyogram signals. To evaluate and compare the proposed model, the common method of artificial neural network (ANN) was used to estimate the force. sEMG signals record from the quadriceps and hamstrings muscles, during isometric muscle contractions (knee extension and flexion) that is done by FUM-PHYSIO robot, and also corresponding force is measured by a force sensor at the same time. these signals are used respectively as the input and the target data in training the SVR and ANN models. Finally, these models tested on healthy people and evaluated by the values of root mean square error (RMSE) and correlation coefficient (CC) between the predicted and the measured force. The results indicate that both methods SVR and ANN have a good performance in estimating muscle force, but the SVR model is more generalizable, more accurate and quicker than the neural network. ε-SVR model also can be seen is more accurate and faster than υ-SVR.