Abstract: When using the time domain matching method to measure the attenuation factor t^* , most of the existing studies rely on the method of artificial vision to screen the benchmark waveform. This process lacks quantitative criteria, is susceptible to subjective bias, and has certain limitations. Based on the fact that teleseismic waveforms recorded by multiple stations in the same network share nearly identical propagation paths, we propose a multi-level screening approach to progressively extract benchmark waveform with more detailed features from waveform records. By combining time-domain matching method with the multi-level screening approach, this paper further explores and refines the measurement process of the attenuation factor t^* of the teleseismic waveform. Numerical experiments show that when t^* is an excessively large positive value, the attenuated waveform becomes overly smoothed, with nearly all detailed features disappearing. Conversely, when t^* is an excessively small negative value, high-frequency noise obscures the dominant characteristics of the waveform, making it difficult to identify its intrinsic properties. Using observational data, this study validates the rationality and effectiveness of the multi-level screening approach by comparing it with previous results and testing different combinations of benchmark waveforms.