In this paper, a high resolution Direction of Arrival (DOA) estimation algorithm based on Khatri-Rao product and PARAFAC decomposition was proposed for a linear array time-delay underdetermined hybrid model of quasi-stationary signals. First, the algebraic structure of quasi stationary signal is used to form the Khatri-Rao(KR) subspace underdetermined blind identification problem. Then, the number of sources and initial value for iteration are estimated using subspace method for improving the estimation accuracy and iteration speed. Finally, the optimal search step linear search iterative least squares method is applied to realize the PARAFAC decomposition, and achieving the DOA estimation of multiple signals. The experimental results show that the proposed algorithm can achieve the higher resolution DOA estimation in the underdetermined hybrid model, especially in the case of low SNR and multi-target condition which has closing incoming directions.