In the near-infrared region, the noise of optical and electronic components affects the signal-to-noise ratio of the second-harmonic signal when wavelength modulation spectroscopy is used to detect gas concentration. Therefore, to suppress the noise, a compound denoising algorithm based on empirical mode decomposition, detrended fluctuation analysis, and the wavelet adaptive threshold is proposed. The algorithm, in view of the existing traditional empirical mode decomposition noise reduction algorithm, is useful for the signal loss problem. Therefore, detrend fluctuation analysis is used to optimize the information intrinsic mode function screening, and it filters out information for the intrinsic mode function for signal reconstruction. Then, the wavelet adaptive threshold algorithm is used to improve the noise reduction accuracy. The proposed algorithm is compared with the classical denoising method and evaluated. The correlation number between the proposed denoising method and the original second-harmonic signal after denoising is 99.9018%, and the root mean square error is 0.0087%. By denoising the second-harmonic signals obtained in the experiment, the results show that the proposed algorithm has a noticeable denoising effect and can retain useful information points.