Interference Separation Using Improved Nonnegative Matrix Factorization for Odor Reproduction Analysis in Mass Spectrum Space

Odorant sample separation is significant for industrial applications and engineering technology. Nowadays, a diverse range of odorant samples is available, including both pure samples and those mixed with chemical compounds, known as fixatives, to extend their shelf-life. Those chemical compounds can cause contradictions when performing odorant analysis with instruments and human noses. This study proposes an improved nonnegative matrix factorization (NMF) to achieve the separation of interference based on a semi-supervised principle to solve the problem. Artificially generated 180 mixed odorants were utilized to calculate odor reproduction analysis. The experiment results show that the proposed method can approximate target odor with appropriate accuracy in numerical results and in actual approximated samples.