The Khatri-Rao subspace methods utilize the characteristics of the Khatri-Rao product of the array response to achieve the extension of the physical aperture and further resolve the quasi-stationary signals in the underdetermined cases. However, the estimation performance of the existing Khatri-Rao subspace methods suffers from the input signal-to-noise (SNR). Especially for a low SNR, the estimation accuracy of the local covariance matrix signals decreases, which will further result in the DOA estimation performance degradation. To address these problems, an enhanced Khatri-Rao subspace method based on beamspace processing for the DOA estimation is proposed. The total array is first split into several identical subarrays, and the beamspace transformation for the data received by subarrays is implemented via beamforming. This results in the improvement of the SNR and further improves the noise suppression ability for the low SNR regions. Then, the Khatri-Rao subspace processing is used for the separated beamspace sources to realize the aperture extension. Finally, the DOAs for the separated sources in the beamspace domain can be further obtained when combined with the subspace-based methods, and the undetermined problem can also be achieved by combining the DOAs obtained from different directions in the beamspace domain.