In order to exactly identify the rotational direction of the measured body in the remote angular-velocity measurement with spatial filtering technique, an approach to estimating the rotational direction was proposed based on the analysis of power spectrum density. In the identification method of rotational direction, two identical photovoltaic-cell arrays with the duty ratio of 0.5 are arranged each other with the shift of a half spatial period, which leads to a differential spatial filter. Two parallel differential spatial filters are placed in a certain interval, and then construct a sensor with double differential spatial filters. The power spectrum densities of the two output signals from one differential spatial filter are conjugated each other, thus the direction coefficients D1 and D2 of the linear velocities on two differential spatial filters are calculated with four output signals of the sensor, and the signs of D1 and D2 represent the moving directions of the measured-body image on the differential spatial filters. The rotational direction and center of the measured-body image can be determined by the signs of D1 and D2, together with the relation between the central frequencies f1 and f2 of two output signals from the differential spatial filters. D1>0 and D2<0 represent the clockwise rotation, and D1<0 and D2>0 represent the counter-clockwise one. For other cases as D1D2>0, the rotational directions can be decided by the sign of D1 and the relative relation between f1 and f2. The proposed method can effectively identify the rotational direction in the remote angular-velocity measurement with spatial filtering technique, which has the features as being single and easy to operate, and having strong adaptability.