The fixed noise covariance matrix fixed by the traditional extended Kalman filter (EKF) could not accurately satisfy the estimation requirement under dynamic and static conditions when observing the rotational speed of induction motor. Aiming at this problem, a fuzzy adaptive method was proposed. The method could adjust the noise covariance matrix parameters according to the state discriminator output, and solve the problem that the fixed noise covariance matrix could not acquire different parameters in different states at the same time. Simulation showed that the fuzzy adaptive EKF had higher accuracy and improved the antiinterference ability of the system.