Abstract: Compared to Finite Impulse Response (FIR) filters, Infinite Impulse Response (ⅡR) filters can match the system better with much fewer coefficients, and hence the computation Load is saved and the performance improves. Therefore, it is attractive to use ⅡR filters instead of FIR filters in Active Noise Control (ANC). However, Filtered-U LMS (FULMS) algorithm, the ⅡR filter-based algorithm commonly used so far cannot ensure global convergence. A new ⅡR filter based adaptive algorithm, which can ensure global convergence with computation load only slightly increasing, is proposed. The algorithm is called Filtered-E LMS (FELMS) algorithm since the error signal is filtered by a time-variant FIR filter. Simulation results show that the FELMS algorithm presents better performance than the FULMS algorithm.