Study On De-noising Of Chromatographic Signal Based On Wavelet Transform

Chromatography is a much faster and more extensive analysis separation technique in analytical chemistry, nowdays. If there is noise in chromatographic signal, it will be difficult to distinguish exactly peak jumping-off point, peak end-point, retention time and peak height , calculate peak area and finally have influence on the quantitative and qualitative analysis in chromatography. So it is very significative to improve the detection limit and the.accuracy and precision of analytical result.The paper introduces wavelet transform, the de-noise theory based on wavelet and its realization in MATLAB software. Then on the basis of traits of chromatographic signals, the evaluating standards, such as peak area relative deviation, peak height relative deviation, retention time relative deviation and signal to noise ration are put forward and the method of de-noising of chromatographic signal based wavelet transform is established.In the process of method establishment, eight waves, seven decomposed levels, four threshold estimators and two threshold functions were chosen by simulated chromatogram, according to evaluated standard. Through a large number of experimental data comparison, the noise filtering method which consists of wavelet db4, decomposition level to 5 or 6, the threshold algorithm heursure and the threshold function soft threshold value is chose to filter noise. Tested by using simulated chromatogram, the results show that the method has a serial of virtues, such as improving the signal to noise ratio, having smaller distortion after filtering noise, and even when the signal has high noise the relative deviation of the peak area ,peak height and retention time are less than 5%, 5% and 1% respectively.Meanwhile, the results of a lot of tests using actual chromatography data, such as capsanthin, rhubarb, Schisandra, and some other chromatographys to filter noise prove that the noise in the original singal could be cleanly eliminated, the characteristics of the signal could be preserved completely ,the position and height of peaks is very clear and it is useful to do quantitative qualitative analysis.In summary, the method is very simple and rapid, which has no restriction on after signals processed, but needs only one filtering to meet processing requirements, and it shortens the processing time.