Resistance spot welding (RSW) is widely used in auto-body manufacturing, and the weld quality is critical to ensure the safety and reliability of an automobile structure. The dynamic resistance (DR) signal is one of the most frequently applied process signals for real-time quality monitoring and control of RSW. However, the accuracy of DR measurement is highly affected by the inductive component of the secondary voltage signal. In a mediumfrequency direct-current (MFDC) RSW system, it is proven that traditional DR measurement methods such as the root mean square (RMS) method and average method fail to eliminate the interference of inductance, and barely meet the monitoring requirements. Therefore, a novel real-time DR measurement method based on a forgetting-factor recursive least square algorithm is proposed and tested under simulated and experimental conditions in this paper. Sensitivity analysis is performed to optimize the forgetting factor, which is proven to be the critical parameter of the novel method to balance the convergence speed and oscillation amplitude. With the optimized forgetting factor, the error analysis and comparative study are conducted under different welding modes. The results show that the inductive noise can be effectively eliminated by the new method. The measurement error of the proposed method is limited within ±6 µΩ at a 99.73% confidence level (±3σ) at both constant and time-varying current modes. This study can pave the way for real-time monitoring and control for an MFDC RSW system.