Purpose - This paper presents the results of a heading estimation method for a remotely operated vehicle (ROV). The output rate of commercially available underwater compasses is typically in the order of a few Hz. Heading frequencies of at least 1 KHz are desirable for navigation and control purposes. Design/methodology/approach - The estimation was performed by fusioning the signals of three inertial sensors: the ROV's own underwater compass (which operates roughly at 10 Hz or less), the ROV's embedded gyro and an additional angular rate sensor that provides readings from 1 to 3 KHz. The output signal of the additional angular rate sensor is not part of the proposed Kalman filter. Nonetheless a five-point Newton-Cotes closed integration of such signal is fed into the Kalman filter implementation that performs the required heading estimation at 1 KHz or more. Findings - The proposed Kalman filter implementation is a suitable approach to estimate heading position even though the original compass signal rate is significantly slower than the signal required for both assisted and autonomous control. Research limitations/ implications - The estimated heading yield good results in both simulation and experimental environments. Originality/value - The method was embedded in a dedicated 16-bit DSP that handles both the acquisition of the three signals and the heading estimation, hence resulting in a very low-cost solution. The embedded solution was tested in the developed submarine and the obtained high-rate heading parameter is now used by the control system of the ROV.