In this paper, a speed sensorless direct torque control strategy for a brushless doubly-fed induction machine (BDFIM) based on extended Kalman filter (EKF) is proposed. By analyzing the mathematical model of a three-phase BDFIM under a rotor speed coordinate system, the state and observation equations of the system are derived, thus realizing the estimations of rotor speed, power, and control winding flux linkage based on the EKF algorithm. According to the estimation of flux linkage, the magnitude of electromagnetic torque is calculated, and the direct torque control of BDFIM is realized by determining the orientation of control winding flux and the inverter’s switching state. Simulation results show that when a step change in speed and a sudden change in load occur, the proposed control strategy can achieve accurate estimations of speed and flux linkage, and the system can run stably.