In order to quantify sound color of rustling sounds of fabrics and investigate the characteristics of specimens classified by sound color factors through cluster analysis, rustling sounds of selected 65 commercial fabrics were produced by the sound generator developed for this study. ΔL, Δf, and LPT values were calculated form sound spectra forms presented by FFT analysis. Specimens were classified into four clusters with ΔL, Δf, and LPT by use of cluster analysis. Means for ΔL and Δf showed significant differences among clusters. Specimens of cluster 1 had higher ΔL and Δf than those of specimens of other cluster. Specimens of cluster 3 lower ΔL and Δf. These indicated that specimen of cluster 1 has deep and less steep curve in sound spectra and those of cluster 3 had shallow and steep curves. It was supposed that specimens of cluster 1 and 3 would generate sounds of unlike color due to the differences of ΔL and Δf.