To study the signal transmission mechanism in the human body, the channel characteristics are generally analyzed by modeling. In current modeling methods, the human body is considered quasi-static and the human tissues isotropic, for simplifying the model and its calculation; however, this does not consider the effect of the human tissues on electric signal transmission, resulting in considerable deviations between the calculated results and the measured values. To reduce model errors and improve precision, a channel modeling method with human muscular-tissue characteristics is proposed in this study. In this method, Maxwell's equations is used as the governing equation and a galvanic-coupling intra-body communication channel model with human-tissue characteristics is built in the cylindrical coordinate system. By building a numerical model with the same parameters as in the analytical model, the analytical solution is proved to be correct. By comparing the different-sample anisotropic models and the isotropic models with the experimental results, it is concluded that the anisotropic model with muscular-tissue characteristics is superior to the isotropic model without muscular-tissue characteristics, with respect to the curve variation tendency and error between the model calculations and the experimental results. The precision of this anisotropic model is enhanced by 200%; hence, it is more accurate. At last, in order to study the optimal communication frequency of the channel, we select 50 healthy persons as the subjects of this experiment, we find that the optimal communication frequency band of the human arm is 10 kHz to 50 kHz. Within this frequency band, the channel gain is the largest, and the mean deviation of samples is less than 2dB, which is very beneficial to signal transmission in human body.