Terahertz time-domain spectroscopy is an effective technique that applies spectral measurements by using an ultra fast femtosecond laser. These unique properties of terahertz waves make it an effective nondestructive testing technology. Currently, it has been applied in many fields. However, when we calculate the optical parameters of the sample especially for liquids, from data obtained with time-domain spectroscopy, a common problem occurred is spurious oscillation, which is caused by the Fabry-Perot effect in sample, containers and optical elements. In order to obtain high frequency resolution and accurate sample optical parameters, it is necessary to use some signal processing techniques properly. In this paper, we improved de-convolution algorithm and presented a simple algorithm based on the traditional optical parameter calculation model. Considering the nonlinear absorption of THz wave by samples, containers, emitters or detectors, a THz time-domain trace containing echo signals can be represented as a convolution of the primary peak, some delta functions and nonlinear transfer functions. By analyzing equations, spurious oscillations in the THz spectra result from echo signals can be removed or reduced effectively. The use of the method is discussed and the transmission spectrum of soybean oil in 0.2~2.0 THz range is measured accurately by terahertz time-domain spectroscopy. Experiments show that, this method can effectively remove the frequency oscillation caused by echoes. This algorithm is not affected by the tested object, and is equally applicable to other liquid measurements in THz range. What’s more, compared with the traditional method of main pulse interception, this algorithm under the same experimental conditions can effectively improve frequency resolution of soybean oil in 0.2~2.0 THz region, from 50 GHz to less than 10 GHz. At the end of this paper, the causation of remaining oscillation in sample spectrum and absorption spectrum were deeply analyzed.