This paper presents a new adaptive scheme for online estimation of stator resistance in speed-sensorless induction motor drives. The method is based on the adaptive control theory of model reference adaptive system (MRAS) approach with Luenberger observer. And the stability of the observer with stator resistance estimation in sensorless vector control of induction motors is proved by the Lyapunov's theorem. Furthermore, the method has been compared with the classical adaptive scheme based on the integration algorithm. And in fact, the proposed method is a generalization of the integration adaptive algorithm. Finally the feasibility of the scheme is verified by simulation