Classification and identification of molecules through factor analysis method based on terahertz spectroscopy.

By means of factor analysis approach, a method of molecule classification is built based on the measured terahertz absorption spectra of the molecules. A data matrix can be obtained by sampling the absorption spectra at different frequency points. The data matrix is then decomposed into the product of two matrices: a weight matrix and a characteristic matrix. By using the K-means clustering to deal with the weight matrix, these molecules can be classified. A group of samples (spirobenzopyran, indole, styrene derivatives and inorganic salts) has been prepared, and measured via a terahertz time-domain spectrometer. These samples are classified with 75% accuracy compared to that directly classified via their molecular formulas.