To overcome the problem that flexible state estimation requires the installation of additional special sensors on hypersonic flight vehicle (HFV). In this paper, a new method is put forward to estimate rigid/flexible states using the Inertial Navigation System (INS) and the Flush Air Data Sensing (FADS) without the installation of additional gyroscopes. The method uses neural network and navigation algorithm to calculate the measured values of FADS and INS to provide low accuracy rigid states and local angle of attack (AOA), the Square Root Cubature Kalman Filter (SRCKF) is then used to fuse the non-linear model of the HFV, the low-precision rigid states and the local AOA to obtain high-precision rigid/flexible states. The analysis results show that the proposed algorithm can maintain high estimation accuracy and fast convergence speed in the presence of model and measurement errors.