Channel modeling is the starting point of effective, efficient body-centric communications. Specifically, path loss prediction plays an important role in estimation of received signal strength, interference optimization and analysis, link budget design and analysis, and cell size estimation. Path loss models are classified as deterministic, empirical, and semi-empirical. The objective of this paper is to estimate path loss model based on measurements of off-body and on-body communications using textile antennas at 2.45 GHz band, and also to analyze propagation channel characteristics. Specifically, model that used in this research is semi-empirical path loss model. All of the used model show the RMSE values are far below 6 dB. This paper also performs propagation channel model estimation from every single modelled scenario. The approach is conducted by comparing the compatibility of path loss distribution with several channel distribution model. These models consist of Log-normal distribution model, Nakagami distribution model, Rayleigh distribution model, and Rician distribution model. Overall, as for off-body communication, Rician and Nakagami distribution model gives equally good representation of measurement data. On the other side, for on-body communication, Rician CDF gives a better fitness to the data than Nakagami CDF.