Abstract: Using modern time-frequency analysis methods: wavelet transform, S-transform, Hilbert-Huang transform and adaptive Chirplet transform, time-varying spectrum estimation was done on Kobe accelerogram record and El Centro accelerogram record. By comparing the analyses, it is showed that the Hilbert-Huang transform and Chirplet transform with the form of adaptive decomposition rules out a variety of interferences caused by the window functions and cross-term and has a high time-frequency resolution; S-transform has low-resolution to a low-frequency component; Although the wavelet transform exists energy leakage in a high-frequency band, it has good temporal resolution. By comparing the time-varying spectrum of the Kobe accelerogram record and El Centro accelerogram record, it is concluded that the Kobe accelerogram record has short duration and high energy characteristics compared with the El Centro accelerogram record, and it could get a clearer time-frequency characteristics of local details by HHT three-dimensional map to understand the distribution of the sequence of energy in terms of a time-frequency plane.