Site amplification of the strong motion of earthquakes is a subject of considerable interest. However, other data are required to produce site amplification factors when available strong motion data are lacking, and this application has been proven to be reliable and practicable. Seismic array observations now cover almost the entire Chinese continent and supply abundant data. Our method is applied in the North China array. Assuming that the seismic stations in a network or array are driven by a forcing function derived from the ambient noise field, each site will respond differently depending upon the elastic and anelastic properties of the underlying medium. With the Steven-Taylor method, the F-K spectrum of ambient noise is regarded as the input and the auto-power spectrum is the response; the ratio between the response and input can thus represent site amplification. In this study, we use the F-K spectrum of the cross-correlation function of ambient noise to deprive the velocity and azimuth of energy extremum, which is different to that in the Steven-Taylor method. We then correct the Fourier spectrum for all stations according to the velocity and azimuth, and the input power spectrum is computed from the stack and average of all the Fourier spectrum. The ratio between the power spectrum of each station and the maximum of frequency-wave number spectrum of ambient noise can then be adopted to estimate site amplification. This method is used to estimate site amplification of the North China array in a period from 10 to 20 s, and the ratio between maximum and minimum amplification is about 7. The sediment thickness of the plain region is basically consistent with that of the Bohai Bay Basin, and site amplification factors of the plain after standardization are found to be larger than those in the mountain region. This study therefore provides an available method to estimate site amplification.