Accurate estimations of unmeasurable variables are required for increasing the performance of sensorless induction motor drives. This can be achieved if the required variables are accurately estimated with a well-established observer.  The paper presents an extended Kalman filter- based direct torque control to investigate the estimations of speed and torque under challenging rotor and stator resistor and load variations at low and high speed regions. In all investigated scenarios, the speed and torque estimation showed good robustness against perturbations and their errors have remained within acceptable error bands.