[A new de-noising technique for spectra based on Mexican hat wavelet]

Signals in spectral analysis often have random noise, which has negative influence on the accuracy and detection limit of analysis. A new chemometrics method named Mexican Hat Wavelet De-noising Arithmetic (MWDA) is presented, which can be used to remove noise in analytical chemical signals. In this method, Mexican Hat wavelet is chosen to construct de-noising function because of its excellent properties, then the de-noising function is used to extract useful information from noisy signals. MWDA is effective for signals with either wide peaks or very sharp peaks. Many processing results of simulated and experimental signals indicate that MWDA is a simple and powerful de-noising method, even when the signal has very high noise (whose signal to noise ratio is 1). After processed, the relative errors of peak position, peak height and peak area are less than 0.2%, 3.2% and 1.1% respectively. When it is applied to experimental spectra, the results are also satisfactory. This new method can increase the accuracy of spectral analysis, and the result is credible and satisfying.