Use of Adaptive Window Wavelet Neural Networks to Solve Inverse Problems of Spectroscopy

摘要

Wavelet transformation uses a special basis widely known for its unique properties, the most important of which are its compactness and multi-resolution analysis of original signal. However, for a standard discrete and continuous wavelet transform (CWT), the extracted set of feature may be not optimal for solving given inverse problem. If no inverse transformation is needed, the values of transition and dilation coefficients may be determined during network training, and the windows corresponding to various wavelet functions may overlap. In this study, we suggest Adaptive Window Wavelet Neural Network (AWWNN) with bottom to top strategy of iterative neighboring windows merging, designed primarily for signal processing. The efficiency of proposed algorithm was compared on the example of the inverse problem (IP) of Raman spectroscopy of complex solutions of inorganic salts. The IP was solved using a dense neural network based on features generated using the proposed approach and a standard CWT.