Analysis of Model Based Virtual Sensing Techniques for Active Noise Control

The performance of active noise controllers based on adaptive filters like the Filtered Reference Least Mean Square algorithm (FxLMS) is optimal in small zones around the error sensor locations. These locations provide the maximal possible reduction of noise but are not accessible by people due to the presence of the sensors. Virtual sensing algorithms can be applied to move the optimal zone of control away from the error sensors. Such methods have been investigated during the last three decades and the most of them rely on initial transfer-function estimation with physical sensors in the virtual locations. This paper investigates how additional physical knowledge about the inherent physics of an active noise control application can be used to derive models, that can extrapolate an arbitrary number of e.g. virtual error sensor signals. A denser grid of error sensors leads to a more homogeneous reduction of noise in a target area and can extend the frequency range of controllability to higher frequencies. The idea of the model based remote microphone technique (MBRMT) is introduced, which is the motivation for this study, and three models for sound-field extrapolation are investigated that could be integrated into the MBRMT.