Polymer flooding significantly enhances crude oil recovery, but its complex rheological properties pose challenges to traditional water holdup measurement. We conduct experimental studies in a 20-mm vertical upward pipe. this work studies ultrasonic propagation in polymer-containing oil-water flows and evaluates different water holdup prediction models. Experiments were conducted using industrial white oil and 200 ppm and 1000 ppm Non-ionic polyacrylamide solutions (NIPAS), covering flow velocities of 5–12 m/day and water cuts of 10%–98%. Signals were synchronously acquired via high-speed camera and ultrasonic sensors, identifying five flow patterns: very fine dispersed oil in water (VFD O/W), dispersed oil in water (D O/W), dispersed large oil bubble in water (D LO/W), dispersed oil slug (D OS/W) and dispersed elongated oil slug (D EOS/W). Combined Weber number (We) and Ohnesorge number (Oh) analysis revealed that viscous forces dominate flow pattern transition at high polymer concentrations. The proposed improved interfacial momentum transfer (IIMT) model introduces a time-harmonic vibration assumption, transforms the dispersed-phase force balance into the frequency domain, and derives the equivalent mixture density through control-volume momentum analysis, establishing a physical relationship between sound velocity and water holdup. This framework effectively bridges microscopic droplet oscillations with the macroscopic acoustic response, providing a unified prediction approach for complex flow patterns. Experimental results demonstrate the high predictive accuracy of the IIMT model, achieving an average absolute deviation (AAD) of 0.0154 and an average absolute percentage deviation (AAPD) of 2.49% at 200 ppm, and an AAD of 0.0145 with an AAPD of 2.06% at 1000 ppm. This performance robustly validates the effectiveness and innovation of the interfacial momentum transfer model proposed in this study.