Aiming at the problem that the dynamic estimation of traditional volume Kalman filter algorithm in coherent optical orthogonal frequency division multiplexing (CO-OFDM) systems is not enough to obtain the optimal phase noise compensation effect, an adaptive cubature Kalman filter(CKF) algorithm based on the block sub-symbol second-order Hermite interpolation (BSSHI) is proposed. The algorithm first estimates the phase noise values of the time-domain OFDM symbol by using the pilot frequency to perform the first second-order Hermite interpolation, and then each OFDM symbol is divided into several sub-symbols to perform the second second-order Hermite interpolation, and the compensation accuracy is improved by adaptively updating the process noise covariance matrix in the CKF. The simulation demonstrate the superior performance of the proposed algorithm in phase noise compensation compared to similar algorithms, and the robustness of the system is improved by adaptively updating the process noise. Therefore, the proposed algorithm can effectively improve the system performance.