Abstract: The signal subspace fitting technique is extended to a general polarization sensitive array for direction-of-arrival (DOA) and polarization estimation of coherent signals. Based on the linear relation between the spatial phase matrix and polarization vector, the new method enables uncouple DOA and polarization estimation via an appropriate separation of signal-subspace rotation matrix. Unlike spatial smoothing and polarimetric smoothing, this generalized signal subspace fitting method (GSSF) does not require a certain number of matched subarrays and can be implemented without aperture reduction. It is shown that GSSF is superior to both the spatial smoothing and polarimetric smoothing preprocessed methods in terms of DOA estimation accuracy, especially in the presence of low signal-to-noise ratio values and short data samples.