Feature Recognition Of Mass Spectrometry Data Based On Continuous Wavelet Transform

As the various kinds of microorganisms, in the field of clinical testing environmental monitoring,food hygiene and other fields,we often need to quickly identify the kind of different microorganisms.Using the feature recognition of mass spectrometry data to identify the different kinds of microorganisms was an very important method.Generally speaking, Using the MS datas to identify the kind of microorganisms usually included the following steps:smoothing denoising, baseline removal, peak extraction, peak position determination, quantitative analysis, etc. The Kalman filter algorithm is simple,saving memory.This paper used Kalman to preprocess the original datas. Continuous wavelet transform algorithm could be quickly realized,has remarkable self-adaptability and the local analysis ability,and the baseline clearance was not required. We could seek peak directly from the filtered datas,this method was better to maintain the characteristics of original datas. Therefore, this paper proposed an improved method of using ridges to search the peaks which was based on the continuous wavelet transform.First we used the Kalman algorithm to smooth the ms datas,and chosed the mexico hat function as an mother wavelet, Then we got the continuous wavelet transform maxima matrix, Identified the ridge lines of maxima matrix and constracted the number of ridge lines.Finally we used the relationship between the ridge lines and the peaks to identify the peaks.The results of the experiment showed that the Kalman filter algorithm could improve the SNR of the original ms datas, and the improved algorithm based on CWT was used to find the peaks by adding a parameter which is called window resolution to bound the number of ridge lines, could better to filtrate the noise of the peaks around the high amplitude, and the number of peaks was in the range of constraints. This method improved the reliability and efficiency of feature recognition Finally, we compared with the feature recognition method of second order derivative, the feature recognition method based on the CWT could quickly and effectively identify the 10 kinds of different microorganisms, the accuracy rate was 100%.