Terahertz time-domain spectroscopy (THz-TDS) allows broadband noninvasive measurement of the optical parameters of various materials in the THz domain. The measurement accuracy of these parameters is highly influenced by the difficulty in distinguishing THz signals from unwanted signals such as noise, signal fluctuation, and multiple echoes, which directly affects material identification and characterization efficiency. We introduce a novel method that provides effective extraction and separation of THz signals from such undesired effects. The proposed algorithm was assessed through experiments that presented enhancement in material parameter evaluation, such as the decomposition of the sample-induced echoes (SIEs) from the complex THz sample signal with near-zero extraction error. Improved precision (+/- 0.05 mu m) was achieved in the determination of the sample thickness compared to that of the mechanical method (+/- 10 mu m). Furthermore, we could infer from the component concentration measurement results of a compound sample (44.2% decrease in the root mean square concentration error) that the material parameter calculation accuracy had improved, proposing a means to enhance the ultimate non-destructive material evaluation performance.