Infrared spectroscopy technology is a new and high technology used in food health inspection, biological pharmacy and environmental monitoring. In order to eliminate baseline drift during application, an infrared spectral baseline correction algorithm based on wavelet optimal decomposition scale is proposed. First, the multi-layer wavelet decomposition and reconstruction are performed on the original spectral signal, and the signal-to-noise ratio of each layer is calculated. Then, the denoised spectral signal is obtained by SNR contrast. The multi-layer wavelet decomposition is performed on the denoised signal to obtain the frequencies of the wavelet detail and wavelet approximation. The two frequencies are divided and the ratio is calculated. Then, the ratios of all the layers are compared, and the maximum one is selected as the best number of decomposition layers. Finally, the spectral signal after baseline correction can be obtained by setting the wavelet approximation coefficient to zero at the optimal decomposition layer number and reconstructing the wavelet. After the experiment verification， it is found that this algorithm can not only provide the basis for the optimal number of layers for wavelet decomposition, but also can remove the high frequency noise and low frequency baseline interference while retaining the useful signals better. The baseline correction is quite sufficient and the effect is good.