Abstract: Based on real-valued decomposition technique, the unitary cyclic MUSIC algorithm is proposed for multipath environment of mobile communication by reconstructing a data model of the cyclic convariance matrix with Hermitian property. Then, a novel cyclic convariance matrix with centro-Hermitian property is designed via spatial smoothing. Thus, a favorable performance is obtained in the presence of multipath propagation. In addition, this approach is available to reduce the computational complexity by real-valued eigendecomposition and be allowed to select desired signals. The simulation results showed that this algorithm performs better not only for independent signals but also for the multipath environment covering coherent signals with lower computational complexity in-comparison with cyclic MUSIC algorithm.