<p>This paper presents an improved state and parameter estimation approach for sensorless control of doubly-fed induction generators (DFIGs) using a modified extended Kalman filter (MEKF). DFIGs benefit from rotor-side measurements, offering greater estimation accuracy compared to squirrel cage induction machines. However, switching noise generated by rotor-side power electronic converters introduces significant disturbances into the measurement signals, degrading observer performance. The proposed MEKF algorithm simultaneously estimates the rotor mechanical speed, stator and rotor flux components, stator/rotor resistances, and load torque in real-time using only current measurements. A core innovation of the approach is the integration of an adaptive noise covariance tuning mechanism. The system noise covariance matrix (Q) is dynamically updated based on operating conditions, while the measurement noise covariance matrix (R) is adjusted using a time- and condition-dependent modeling technique to effectively mitigate the impact of switching noise. In contrast to hybrid or structure-switching observers, the proposed method captures frequency- and temperature-dependent resistance variations within a single MEKF framework without additional observer switching logic. This results in improved estimation robustness and enhanced control performance under variable operating conditions. Simulation results in MATLAB confirm the effectiveness of the proposed method. The adaptive MEKF achieves significantly higher estimation accuracy than conventional EKF and MEKF implementations that do not account for switching noise. These findings demonstrate the potential of the proposed observer for reliable sensorless motion control in DFIG-based wind energy systems.</p>