A Novel Peak Detection and Data Fusion Methodology for Multidimensional Chemical Analysis

Intelligent Automation, Inc. (IAI) proposes a novel peak detection and data fusion methodology for multidimensional chemical analysis. The proposed peak detection technique is based on Continuous Wavelet Transform (CWT). Its distinction from other peak detection methods is that it does not require any baseline removal or smoothing algorithms and is robust to noise. It can differentiate the signal from the spike and colored noise and the signal-to-noise ratio can be greatly enhanced. This results in significant reduction in false alarm rates. Since the number of the detected peaks in a multidimensional space may be excessively large, efficient data reduction techniques are needed so that efficient data classification techniques can be applied to a low dimensional data. Our proposed feature extraction method takes advantage of the theoretically well-founded and established data dimensionality reduction techniques. Development of efficient data classification and data fusion are two other important components in the proposed effort. We propose to develop a Maximal Entropy (ME)-based data classification approach. Different from the existing the classifiers the proposed ME-based classification approach can automatically identify unknown chemical agents of which their signatures do not exist in the database. The classification results from different multidimensional chemical analysis techniques will then be fused by an ME-based fusion algorithm. The fusion output will be a set of probabilities of the suspected material belonging to certain kinds of known chemical agents or hazardous materials.