ISO/IEC 19794-2 compliant fingerprint minutiae template is an unordered and variable-size point set data. Such characteristic leads to restriction to the applications that can only operate on the ordered fixed-length bit-string, such as cryptographic protocols and biometric cryptosystem scheme like fuzzy commitment and fuzzy extractor operating in hamming domain. In this paper, we propose a discriminative fixed-length binary representation converted from fingerprint minutia based on Kernelized Locality-Sensitive Hashing (KLSH), which enables speedy matching. The proposed method includes four steps: minutiae descriptor extraction; Kernelized Locality-Sensitive Hashing for fixed length vector generation; dynamic feature binarization and matching. Experimental results on FVC2002 databases justify the feasibility of the proposed template in terms of matching accuracy and template randomness.