Application of near-infrared spectroscopy to the prediction of sample content is strongly limited by signal peak overlap. To analyze the spectral information directly related to the target components and to make the chemometric model more explanatory, an independent characteristic projection algorithm is proposed. The algorithm was applied to the independent spectral analysis of a single sample using corn as a representative example. Moisture, oil, protein, and starch, which are the four main components of corn, were the target components. The pure component spectra were used the projection directions to decompose the near-infrared spectrum of a single corn sample; then four decomposed spectra corresponding to the four pure component spectra were obtained. Their corresponding relationship was determined using their correlation coefficients and by comparing their characteristic peaks, and the molecular absorption patterns corresponding to the characteristic absorption peaks of each decomposed spectrum were analyzed in detail. The theoretical analysis and experimental results indicate that the independent characteristic projection algorithm can be applied to single-sample spectral analysis to extract more complete physicochemical information about the target components and provide a theoretical basis for establishing a robust near-infrared spectral chemometric model with great extrapolation capability and stability.