Adapting transfer functions to changes in atmospheric conditions for outdoor sound field control

The performance of active sound field control solutions is directly dependent on the accuracy of the measured acoustic transfer functions between the control loudspeakers and control areas. Outdoors, these transfer functions are affected by atmospheric conditions and their variation over time. In this work, we investigate strategies for transforming measured transfer functions to changing atmospheric conditions with the goal of adapting outdoor sound field control systems to such changes. Compared to active control methods based on adaptive filtering, such weather model based approaches do not rely on continuous sound pressure recordings inside the control area. We investigate different adaption strategies, one based on explicit delay compensation with effective speed of sound and an alternative based on machine learning. We train and test the adaption strategies against results of a large set of outdoor transfer function measurements in different atmospheric conditions.