Abstract: Based on theoretical analyses, this paper indicates that the smaller the distance between the basic wavelet and the analyzed signal (in a sense of L²(R,dt) space, the more concentrated is its energy distribution in the time-scale domain, that is, the less is the dimensions of EDS of the signal, and vice versa. Then the EDS of signals and colored noises are studied, respectively. Based on these works, an interpretation method of seismic data is proposed by using the physical wavelet best matching the seismic wavelet (or significant signal). By this method, the noises and the interference waves are partially suppressed, at the same time, the useful information is enhanced. The efficiency of the method has been proved by modeling and field data.