GNSS (Global Navigation Satellite System) observation is complicated by the presence of noise, local benchmark motion, un-modeled atmospheric and other effects. More subtle signals have existed the individual GNSS time series but have undetected due to its low signal-to-noise ratio (SNR). We introduce here a new method to extract spatially and temporally coherent signals by analyzing data from entire GNSS networks simultaneously using free-net solutions. Daily GNSS network is performed a similarity transformation by constraining one location and one direction. The deformation signal is extracted by analyzing the change of the GNSS geometry net-form integrating the location, baseline length and direction information of all GNSS stations from entire networks. Combined with area strain, the deformation characteristic are further interpreted. We apply the method for the 2013 Lushan Ms 7.0 earthquake and find a significant pre-seismic deformation anomaly with the locked status and negative acceleration in the first shear strain and direction change time series, implying a strong sinistral shear tectonic forces have happened in seismic source area and can intensify the pregnancy of the earthquake. The phenomena can be well supported and explained by the experiment results of rock fracture in the laboratory. It's inferred that a processes from stable locked status to negative acceleration change may be the most characteristic of pre-seismic deformation.