Due to its lower concentration limit, wavelength modulation spectroscopy has been widely used for gas concentration detection based on laser absorption spectroscopy. Various background noises and optical fringes can decrease the accuracy of concentration measurement and increase the lower limit of concentration measurement. To reduce the background noise and improve the signal-to-noise ratio of the measurement system, the absorption spectral lines at 2.25 μm were used to measure the NH₃ concentration in the experiment and theoretically analyze. In signal processing, wavelet and Gabor transform techniques were used to reduce the noise in harmonic signal. The results show that Gabor and wavelet transforms can effectively remove the influence of various noises while preserving the integrity of the harmonic signal with the optimal control parameters. Wavelet transform has better denoising ability than the Gabor transform for harmonic signals mainly containing white noise and optical fringes, and can improve the signal-to-noise ratio and the smoothness of the signal curve. The lower limit of NH₃ concentration after denoising is reduced from 0.36 mg/m³ to 34.45×10^-3 mg/m³.