Underwater time delay estimation is crucial for high-precision localization. However, multipath effects in the underwater acoustic channel cause modal aliasing of received signals in the time-frequency domain. Coupled with the non-stationarity of ocean ambient noise (e.g., ship radiation, biological activity), the performance of existing time delay estimation algorithms is significantly degraded. To address this problem, this paper proposes a time delay estimation method based on spectral entropy-driven multipath separation. Through the spectral entropy-driven bandwidth overlap criterion, effective multipath modes are dynamically selected, and noise-dominated modes are discarded; the time-domain energy detection mechanism is designed to optimize time delay estimation, incorporating both energy gradient and information entropy. The method overcomes the limitations of fixed-parameter modal decomposition and solves the problem of separating time-frequency aliased signals and achieving high-precision time delay estimation under low signal-to-noise ratio. Simulation experiments with linear frequency-modulated and hyperbolic frequency-modulated signals demonstrate that the accuracy of multipath separation is improved 52.1%–61.4% compared to traditional algorithms, and the root mean square error of the time delay is reduced 36.6%–47.2% compared to the time-frequency ridge-tracking method, providing a theoretical breakthrough for multipath localization in a complex underwater acoustic environment.
Cui et al. (Thu,) studied this question.