Pedestrian inertial navigation systems accumulate errors over time during computation, leading to divergent positioning results. The traditional Generalized Likelihood Ratio Test（GLRT） zero-velocity correction algorithm is insufficient to suppress the divergence of positioning errors. To address this limitation, this paper proposes a low-cost, high-precision detection method based on an ultrasonic ranging sensor to assist the GLRT zero-velocity correction （abbreviated as UA-GLRT）. Using a standard basketball court as the experimental site, five independent repeatability experiments were conducted to compare the effectiveness of the UA-GLRT algorithm with that of the GLRT algorithm. The two zero-velocity detection algorithms were evaluated through position errors and loop closure errors. Experimental results demonstrate that the UA-GLRT algorithm reduced the average start-end position error from 1.02 m to 0.43 m and decreased the average loop closure error from 1.19%D to 0.5%D compared to the baseline GLRT method.