Most of current myoelectric prostheses are using sequential and on-off control strategy within pattern classification framework, which is of robustness. But it is not a natural neuromuscular control scheme. On the other hand, there are two difficulties to control the prosthesis proportionally and simultaneously. First, human hand is high dimensional with more than 20 degrees-of-freedom (DOFs); Second, extracting such control information from EMG is hard due to signal crosstalk and noises. This paper is aimed at proposing a new method for proportional and simultaneous myoelectric control, taking advantage of synergy concept. The hand motion and corresponding forearm EMG signals were collected simultaneously. Principal component analysis (PCA) is used to reduce hand motion dimension. And support vector regression (SVR) is adopted to build the connection between hand posture and EMG. Offline analysis validated the effectiveness of this method, and preliminary and positive results have been obtained.