We applied near-infrared spectroscopy with the aid of chemometrics to analyze non-destructively blended fabrics whose components are two or three of eight fibers, i.e. cotton, hemp, rayon, wool, silk, nylon, polyester, and acrylonitrile. Materials identification was achieved by the best-fit reproduction of observed spectra with the linear combination of spectra of fabrics containing a single material and by identification analysis with principal component analysis (PCA) score-plots. Blend ratios were estimated with models constructed by partial least squares (PLS) regression method. Our technique is expected to enable non-destructive and quick analyses of blended fabrics.