Quantitative ultrasound has been suggested for screening of osteoporosis. Most commercial ultrasound devices are based on the through-transmission measurement of calcaneus, which is not a typical fracture site. In contrast to through-transmission measurements, reflection and backscattering measurements may be conducted at typical fracture sites such as vertebra and proximal femur. At these regions, soft tissues overlying bones affect reliability of the measurements. In this study, a novel dual-frequency ultrasound (DFUS) pulse-echo technique is introduced for reduction of the errors induced by soft tissues. First, DFUS was validated using elastomer samples. For further validation, human trabecular bone samples (n = 25) covered with heterogeneous soft tissues were measured at frequencies of 2.25 MHz and 5.0 MHz. The DFUS technique reduced (p < 0.01) the mean error induced by soft tissue from 58.6% to -4.9% and from 127.4% to 23.8% in broadband ultrasound backscattering and integrated reflection coefficient (at 5.0 MHz), respectively. To conclude, the DFUS, being the first ultrasound technique capable of determination of the composition and thickness of the soft tissue overlying the bone, may enhance the accuracy of clinical ultrasound measurements. Thereby, DFUS shows a significant clinical potential.