Principal component analyses (PCA) were conducted for skin spectral reflectance of a new skin colour database. Results for skin colours from different ethnic groups were analysed and reflectance re-construction based on first three components obtained by PCA were investigated. Significant differences in the derived principal components for the three ethnic groups were found. Furthermore, a new nonlinear optimization model based on all reflectance data was developed for generating the three basis functions for reflectance reconstruction. Comparison results for reflectance predictions show the basis functions obtained from the new optimisation model are better than those obtained from PCA.