In order to eliminate the influence of noise in laser absorption spectral signal on analysis result, we propose a threshold denoising method based on wavelet transform. The simulation experiment is implemented with the utilization of direct absorption spectroscopy signal of gas molecules combined with MATLAB software, and the effects of threshold methods, decomposition level and wavelet function type on denoising effect are analyzed in details. Finally, the optimized filtering parameters are used to analyse the spectral signal of human expired gas in experiment. The results show that the wavelet transform algorithm developed in this study has a good denoising effect for the spectral signal. In addition, real-time analysis of the other component of expired gas can be realized by selecting different tunable laser light sources at other wavelength. The quantum cascade laser (QCL) spectroscopy system can be widely used in the field of expired gas diagnosis and other fields.