<p>This paper proposes an improved time-frequency direction of arrival (DOA) estimation algorithm based on accelerated structured alternating projections (ASAP), to address the severe performance degradation of conventional DOA estimation methods under mixed noise and interference in complex electromagnetic environments. First, the low-rank property of the Hankel matrix is exploited to perform structured matrix analysis of the observed signals, and an accelerated low-rank approximation-based subspace projection method is applied through ASAP to suppress noise and interference robustly. Second, an improved synchrosqueezing transform (SST) algorithm is developed for time-frequency analysis, enabling the capture of local time-frequency features of the signals. Finally, an edge-linking algorithm based on local peak detection in the time-frequency representation is used to extract time-frequency trajectories across multiple channels cooperatively and to build a space-time-frequency-weighted covariance matrix, thereby ensuring high-precision DOA estimation. Extensive simulation and experimental data show that the proposed algorithm significantly outperforms existing benchmarks in estimation accuracy, successful detection probability (SDP), and robustness against mixed noise across the entire signal-to-noise ratio (SNR) range. At an SNR of −10 dB, the proposed method achieves a 3.6% to 44.7% reduction in root mean square error (RMSE) and a 5% to 20% improvement in SDP. It also demonstrates satisfactory computational efficiency, with an average runtime of 0.057s per run, indicating its suitability for real-time applications, albeit with increased computational overhead and the need for parameter tuning.</p>