Wavelength selection of amino acid THz absorption spectra for quantitative analysis by a self-adaptive genetic algorithm and comparison with mwPLS

For researchers interested in the quantitative analysis of mixture components by terahertz time-domain spectroscopy, the wavelength selection of the full spectrum obtained by the experiments is of great importance. The full spectrum is not only composed of the mixture components' characteristic features, but also many types of noises, such as the scattering and the instrumental disturbances. Because of this, the accuracy of the quantitative analysis may be very unsatisfactory if the full spectrum is used. Therefore, the full spectrum must be deeply selected, which means retaining the useful information points and dropping the noises. In this paper, a self-adaptive genetic algorithm (sa-GA) was proposed for point-by-point wavelength selection of mixture THz absorption spectra, with the crossover and mutation probabilities dynamically adjusted, which is helpful for improving the method's performance. The sa-GA was compared with the moving window partial least squares (mwPLS) method, which has been proved to be an effective method for wavelength selection in THz spectroscopy. The quantitative analysis experiments of a series of amino acid mixtures proved that the sa-GA was superior to mwPLS, with lower error (1.94 ± 3.79%) and better convergence.