ROBUST FEEDBACK ACTIVE NOISE CONTROL IN HEADPHONES BASED ON A DATA-DRIVEN UNCERTAINTY MODEL

Active noise control (ANC) currently receives academical and practical interest that is accompanied by a rising popularity in consumer devices. Feedback control offers an appealing option to implement ANC but is limited by the uncertainty that is inherent to the controlled system. In this work, we show that the conventional disk-based uncertainty model may overestimate the uncertainty and thus limit ANC performance. We propose a data-driven model that exploits the spectral shape suggested by experimental measurement data as an appealing alternative. We integrate this model into an established controller design algorithm to facilitate a more targeted trade-off between robustness and performance. Through simulation and measurement which involves the implementation of an ANC prototype on a real-time platform, we demonstrate that the use of this model can yield robust controllers with improved performance.