Fingerprint recognition and retrieval is a widely used biometric application for identification and security purpose. Various methods has been proposed for fingerprint identification  and recognition and each methods has advantage as well as disadvantages that makes the system partially efficient. The complexity in process and other issues affects performance of the existing system and make the system inefficient. In this paper we presented a Singular Point Feature Block [SPFB] based segmentation of fingerprint in the generation of finger code for better recognition. The feature set is generated based on the singularity points such as core point, delta point and reference point and the global feature entropy. In the first phase the orientation field is estimated, singular points are detected and segment the image based on block generation around the singularity points. Hence a finger print code is obtained with singular point feature. In the next phase the feature set is generated and matching is done with various fingerprint images and this method produces an efficient result and cost effective. Euclidean distance is used for distance measure between two image features. The images from the standard image data base FVC2004 of DB3 is considered for the experimentation; FRR and FAR has been compute for performance evaluation. This proposed approach is compared with the existing methods and it provides better results