Classification of biological spectrum based on principal component cluster analysis

摘要

Spectrums of 17 biological tissue phantoms were measured using the fiber-optic spectrometer. Then, the spectrum was preprocessed by multiplicative scatter correction method to devoice the spectrum. Afterwards the features of the spectrum were extracted via principal component analysis. Ultimately, we applied cluster analysis for the spectral features. The results showed that the accumulated credibility of the first 12 spectral principal components was 99.86% for the spectrum after preprocessing; indicating that this spectrum feature extraction might be done in the case of losing no key information. And the results showed that the 17 biological tissue phantoms can be divided into four main categories according their optical features.