It has been shown that freestream velocity disturbances may cause substantial aeroelastic and aerothermal loading on a hypersonic flight vehicle due to the inadvertent formation of shock waves and rapid changes in air density. We present here a novel state estimation framework to determine and predict the air density acting on a hypersonic flight vehicle. Our approach is comprised of a learning algorithm that updates the air density estimate given observations made by conventional inertial measurement unit sensors earlier on in flight, while exploiting known maximum bounds on acceleration changes and angular rates in the given flight regime. The primary motivation for this research is to enable predictive maneuvering so as to anticipate density perturbations, thereby alleviating aerothermal loads. Previous related work has seen many applications in the field of low speed aeronautics, relying on Kalman and moving average window filters, which suffice only in the case of low frequency changes, require heuristics in tuning them, and are not applicable to the high frequency flow perturbations that are experienced in hypersonic flight. We demonstrate the proposed approach by applying it to the entry trajectory of the Mars Phoenix lander. The atmospheric properties obtained by applying our methodology to this test case are validated using previous trajectory reconstruction efforts and the NASA Mars-GRAM 2010 atmospheric model, allowing for validation of the estimate atmospheric properties generated by the present algorithm.