In order to address the increased measurement errors and reduced positioning accuracy and reliability caused by factors such as complex channel environments, multipath effects, and signal attenuation in practical applications, this paper proposes an improved method for the Ultra-Short Baseline (USBL) positioning algorithm. The method involves adaptive filtering of the received signals to remove noise and interference and enhance signal quality. Envelope detection is then utilized to extract important feature information from the signals. Finally, fitting technique are applied to process the envelopes of the correlation peaks, resulting in higher time delay estimation accuracy and improved target positioning accuracy. Through simulations and experiments, the performance advantages of the proposed method over traditional algorithm are demonstrated, validating its feasibility and effectiveness in real-world applications.