Surface electromyography signal (sEMG) is the recruitment of electromyography signal on the human skin surface and has become an ideal source of control signals for artificial prostheses. This paper studies the relationship between the characteristics of sEMG and arm muscle strength. Firstly, the collected signal is denoised by using wavelet transform, and secondly the characteristics of the denoised sEMG are analyzed from the time-frequency domain. Then classify sEMG by the support vector machine classifier to establish the relationship between sEMG characteristics and arm muscle strength. Finally, the arm muscle force is converted into current, and establish the relationship between sEMG eigenvalues, arm muscle strength and current, thereby realizing intelligent prosthesis control.