Wavelet Transform and Artificial Neutral Network for the Quantitative Resolution of Ternary Mixtures

Abstract

Two different approaches continuous wavelet transform (CWT) and artificial neural network (ANN) were successfully applied to the quantitative resolution of ternary mixtures of paracetamol (PAR), metamizol (MET) and caffeine (CA) having strongly overlapping spectra. First approach is based on the use of CWT signals of the ratio spectra of three active compounds in samples. Various CWT families were tested for the extraction of cardinal information and small noise condition and higher peaks of the original spectra. In this paper, three methods: Mexican hat function (MEX) (a = 70), reverse biorthogonal (RBIO3.5) (a = 100) and biorthogonal (BIOR2.4) (a = 90) were found suitable for determination of three active compounds and their signal analysis. ANN, as a multivariate numerical method, was used for the quantitative resolution of the same ternary mixtures. The performance of CWT and ANN approaches was validated by analyzing the synthetic mixtures of PAR, MET and CA compounds. In addition, the standard addition technique was also used for the same purpose. The experimental results provided by CWT technique were compared with each other and those obtained by ANN method and a coincidence was observed for all the obtained results.