Abstract: In order to improve the accuracy of direction of arrival (DOA) estimation with arbitrary array, from point of view of optimizing the subarray elements, an optimal subarray partition estimation of signal parameters via rotational invariance technique (ESPRIT) based DOA estimation algorithm with high accuracy for arbitrary linear array is proposed. Firstly, virtual interpolated array estimation of signal parameters via rotational invariance technique (VIA-ESPRIT) is utilized to obtain the coarse estimation. Secondly, to acquire the rotation invariant feature, the coarse estimation serves as the reference to carry out the phase compensation on arbitrary array. The subarrays are optimally partitioned to construct the invariance equation are according to the ESPRIT principle. Then the rotational invariance between the optimized subarrays is utilized to obtain a high accuracy DOA estimation. Furthermore, the effect of subarray partition on the performance of DOA estimation is analyzed, and an approximate calculating method of optimal subarray partition. Computer simulation results show the effectiveness of the proposed algorithm, and its high performance is close to the Cramer-Rao bound (CRB).