A pre-process, feature extraction and quantitative analysis approach based on principal component analysis (PCA) is proposed to analyze the complicated nonlinear fluorescent spectra, emitted by the interaction between femto-second(fs) laser and the impurities in air. The spectral data is denoised and compressed from 3979 to 664 points using wavelet transform (WT). By fitting the feature peaks of the compressed spectra with different concentration impurities using PCA, the monotone relation between intensity and concentration is identified and it can be used to perform quantitative analysis. Simulation results on three kinds of impurities with low concentration show that the first two principal components can cover 98% information, and the quantitative analysis method based on the first principal component can effectively reduce the error from 0.2694 to 0.02 compared with previous method.