Stoke causes dysfunction of limb and seriously affects the life quality of patients. In the upper limb rehabilitation, doctors need to carry out subjective  assessment for upper-limb of patients. But this method has large error and high cost. Therefore, artificial intelligence technology is applied to the field of medical rehabilitation. This paper summarize the objective assessment methods which are base on sEMG signal feature, trajectory error feature, joint angels and joint angular velocity. Subjective assessment methods include Brunnstrom , Ueda, Bin , Fugl-Meyer, and Wolf Motor Function Test. Finally, the paper concludes that existing objective methods are generally affected by the scale of data and the number of feature. Subjective methods are generally limited by the time-consuming and subjective error. In the future, we should improve the objective algorithm from accuracy, scale of data and multi-feature fusion.