A fast and nondestructive identification method to distinguish different types of fabric fibers is proposed in the present paper. A total of 214 fabric fiber samples, including wool, cashmere, terylene, polyamide, polyurethane, silk, flax, linen, cotton, viscose, cotton-flax blending, terylene-cotton blending, and wool-cashmere blending, were collected from Beijing Textile Fibre Inspection Institute. They contain yarns, raw wool or cashmere, and various fabric straps with different colors and different braid patterns. Sample presentation for measuring near infrared spectra of various textile fibers was tried to reduce the impact from the ununiformity of polymorphous fabric structure. Spectral data were pretreated using multiplicative signal correction (MSC) to reduce the influence of spectral noise and baseline shift. Classification of 12 kinds of fabric fibers in various braid patterns was studied using minimum spanning tree method and soft independent modeling of class analogy (SIMCA) classification based on principal component analysis of NIR spectra. The minimum spanning tree for the spectra of total samples shows that the samples in the same type fall almost into one cluster, but there are overlaps between some two different clusters of fabric fibers with very similar chemical compositions, such as wool and cashmere. Complete discrimination between cashmere and wool has been achieved using SIMCA. The results show that nondestructive and fast identification of fabric fibers using near infrared spectral technique is potentially feasible.