Multiple signal classification(MUSIC) is a computationally complex algorithm that is hard to implement for real-time application. First, for the case of uniform linear array(ULA), the preprocessing algorithm making the following computations in real-value is presented and a practical spatial spectrum is defined. The algorithms for eigenvalue decomposition(EVD) suitable for hardware implementation is selected as well. Then the overall architecture of a FPGA implementation of MUSIC is proposed. Verification results show that this implementation with FPGA can compute MUSIC accurately with high speed.