Evaluation of the efficiency of amphetamines detection by cluster analysis

In this paper we are evaluating three signal processing methods applied for improving the efficiency of the detection of illicit amphetamines performed with Principal Component Analysis. The GC-FTIR spectra of the targeted compounds have been preprocessed by using two functions acting as selective amplifiers and one as an amplifying selector. The efficiency of the detection system was evaluated for each feature weight by using dendrograms obtained by agglomerative clustering. A final ranking of the preprocessing functions was performed based on the cophenetic correlation coefficient. The results confirm the usefulness of the proposed signal preprocessing methods. © 2014 U.P.B. Sci. Bull.