Abstract: To address the difficulty in identifying the first-arrival path—caused by signal multipath superposition due to boundary reflections in shallow and confined waters—this paper proposes an underwater multipath ranging algorithm based on adaptive threshold decision. The algorithm employs a hyperbolic frequency-modulated (HFM) signal as the ranging excitation, uses overlap-add FFT to compute the correlation of continuously received signals, constructs a joint decision model that integrates a dynamic background-noise threshold and a peak-ratio threshold, and incorporates a continuous local-peak constraint to detect the first-arrival path. Experimental results show that, compared with the basic matched cross-correlation method, the proposed algorithm reduces the mean absolute error by approximately 85.2% in a multipath test pool environment. In a shallow-water experiment conducted in the South China Sea at a range of approximately 800 m, the mean absolute error was 3.41 m and the mean relative error was 0.41%. These results indicate that the proposed algorithm effectively reduces the probability of range estimation errors induced by strong reflection paths, thereby providing technical support for underwater acoustic ranging in complex multipath environments.