Ideal reference sites of the generalized inversion technique (GIT) are not always available. To overcome this limitation, in this article, average site amplification factor, calculated by the quarter wave-length method, and the high-frequency attenuation parameter, obtained from the spectral decay method, were adopted as the empirical reference site (ERS) of the GIT. Rationality of ERS-GIT was verified, through comparison with other methods, based on 615 groups of strong ground motion records of 95 aftershocks of the 2008 Wenchuan M8.0 earthquake, recorded by 28 strong motion stations during May and August 2008. Reasons for the differences among the various methods are discussed. Results showed that the average stress drop, as calculated by the ERS-GIT, classical GIT method, and nonparametric GIT method was 1.15 MPa, 0.78 MPa, and 0.52 MPa, respectively. Quality factors obtained from the results of the ERS-GIT and classical GIT methods were Q(f)=75.02f^1.27 and Q(f)=65.56f^1.22, respectively. Taking the predominant frequency of the site as a boundary, the local site responses calculated by the ERS-GIT coincided with those of the H/V spectral ratio method and classical GIT method at long period and high frequency, respectively.