Optimum Fixed-Pole Adaptive Filtering for Active Noise Control in Headphones

The implementation of least mean square (LMS)-type algorithms for active noise control (ANC) in headphones entails a trade-off between performance and computational complexity. In both aspects, adaptive infinite impulse response (IIR) filters with pre-determined pole locations, commonly referred to as fixed-pole adaptive filters (FPAFs), offer advantages over widely-used finite impulse response (FIR) filters. The optimal selection of the poles requires oracle knowledge, but remains ambiguous under practical conditions. As a remedy, a novel loss based on the steady-state error is proposed. The associated pole location optimization which yields stable low-order FPAFs is described. Using real-world measurement data, the performance and the degree of robustness against transfer path variations are investigated. The results indicate an improvement of several decibels over conventional algorithms with similar implementation costs.