The detection of one or more trace gases in the noise background is difficult, since the absorption signal is weak and there exist various disturbances in the measurement system.Several methods have been developed to deal with this problem based on the absorption spectroscopy of gas, e.g.the differential optical absorption spectroscopy (DOAS) and the wavelength modulation spectroscopy (WMS).However, the accuracy of DOAS is susceptible to the fluctuation of optical power and the presence of other gas in the system.The signal amplitude in the WMS scheme may be influenced by the residual-amplitude-modulation (RAM) effect which will limit the precision of detection.To overcome the interference and random background noise in the measurement system,the authors had proposed an averaging scheme to extract the trace gas signal in time domain.This time domain signal processing method needs a complicated wavelength reference and calibration procedure to ensure the supposition of each scanning has the same starting point.To avoid such complex and time-consuming but essential wavelength calibration work in conventional measurement, we propose a novel averaging algorithm which doing the signal processing in frequency domain.Since each gas possesses a unique comb-like absorption spectrum in the interested wavelength range, so, by doing FFT to the spectrum data obtained from the wavelength scanning measurement, the trace gas can be identified from the particular frequency components in frequency domain.Furthermore, the weak gas signal can be extracted by the multiple suppositions of the FFT results.It shout be pointed out that the proposed averaging algorithm in frequency domain doesn't need a beforehand wavelength calibration procedure, since the FFT amplitude is insensitive to the'phase', i.e.the small wavelength shift in the spectrum data obtained from the wavelength scanning.Therefore, without the accurate calibration unit the measurement system can be greatly simplified and become fast and reliable.In the experiment, a maximum retrieve deviation of 10.6% in the measurement of 100 ppm C2H2 gas sample was demonstrated.The result and analysis show the proposed frequency domain signal processing scheme has good prospect in the application of gas detection.