Adaptive Feedback Active Noise Control by Robust Controller Interpolation

Feedback active noise control (ANC) offers an economical means to attenuate noise and is thus used in a variety of problems, e.g., in automotive or hearables. In practice, stability needs to be guaranteed also in the face of variations in the controlled system. Typically, a constrained optimization problem is solved to integrate this so-called robust stability criterion into the system design. However, the tracking of constraints at runtime remains challenging in resource-constrained environments such as hearables. In this contribution, we propose an efficient algorithm which is suitable for such environments. This algorithm uses internal model control (IMC) to steer a parallel connection of time-invariant filters via uncertainty-aware least-mean-square (LMS) adaptation, thus facilitating the adaptation at high sampling rates. The novel algorithm is evaluated through simulation and measurements with human listeners using ANC headphones as an example use case. The results show good ANC performance despite significant savings.