Performance of discrete wavelet technique in the de-noising of noisy spectrum in the mid-infrared region near 4.5 µm

Abstract

We applied discrete wavelet transform (DWT) technique for de-noising of carbon monoxide (CO) spectrum which is experimentally obtained in the mid-infrared region using a difference-frequency-based spectrometer. The performance of DWT is firstly examined by de-noising of a simulated P(16) CO absorption line based on the real data associated with the experimental setup we arranged in laboratory. It is found that when the db20 function from Daubechie wavelet family is used at level 7 of composition, highest signal-to-noise ratio (SNR) ~23 can be achieved. Similar results are obtained in the experiment in which a noisy CO trace is de-noised using the same procedure as described in the simulation. Subsequently, highest SNR of ~9.2 and lowest residuals is experimentally obtained using the db20 function which confirms the merit of DWT technique in detection of low-level absorption signals.