[Objective] To address the issue that the sensorless control process of interior permanent magnet synchronous motor (IPMSM) is susceptible to magnetic circuit saturation, which induces variations in dq-axis inductance parameters and model mismatch thereby degrading control performance, this paper employs an extended Kalman filter (EKF)-based online identification method to estimate the dq-axis inductances in real time. [Methods] The EKF algorithm was designed based on the optimal recursive estimation principle for nonlinear systems. System and measurement noises were processed through real-time updates of state prediction and the error covariance matrix. The identified inductance parameters were fed forward into the sensorless observer algorithm to achieve dynamic updating of the motor model and closed-loop feedback correction, thereby enhancing the system’s robustness against parameter variations and complex operating conditions. [Results] The experimental results indicated that with the introduction of EKF online parameter identification, favorable dynamic and steady-state performance was still maintained under low-speed loaded conditions; moreover, when a sudden load was applied in the high-speed region, the rotor position observation accuracy was enhanced by 1.5°. [Conclusion] Comparative analysis with the ordinary least squares method confirms that the EKF-based parameter identification demonstrates superior performance, effectively suppressing the impact of magnetic saturation and cross-saturation effects on position estimation.