Tunable diode laser absorption spectroscopy signal denoising based on multi-resolution singular value decomposition and improved complete ensemble empirical mode decomposition with adaptive noise optimized by the improved crested porcupine optimizer algorithm

When utilizing the tunable diode laser absorption spectroscopy technique for gas concentration measurement, the second harmonic signal is significantly affected by noise, leading to a decrease in measurement accuracy. We proposed a noise reduction algorithm to address this issue. It combines a multi-resolution singular value decomposition (MRSVD) and an improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) algorithm optimized by the improved crested porcupine optimizer (ICPO) algorithm. First, the MRSVD algorithm is utilized to denoise the signal, followed by optimizing the ICEEMDAN parameters using the ICPO algorithm. Subsequently, the signal is decomposed by ICEEMDAN and reconstructed based on the correlation coefficients of intrinsic mode functions to obtain a final denoised signal. In the experiment, the proposed algorithm achieved a signal-to-noise ratio of 116.8707 dB, and the linearity between the gas concentration and the maximum value of the second harmonic signal improved from 0.9893 to 0.9984. Moreover, the average relative error of the gas concentration decreased by 3.12%. The proposed algorithm exhibits superior denoising efficiency compared with traditional methods and proves effective for denoising open optical range signals.