Wind speed and direction estimation based on the spatial coherence of closely spaced microphones

Wind-induced noise recorded with a compact microphone array can be exploited to infer the mean velocity of the free-field airflow. In this work, a model-based method to estimate the wind flow speed and direction is proposed that uses spectro-spatial correlations of closely spaced microphone signals. As shown in a recent work by the present authors, the normalized cross-power spectral density of flow-induced noise measured with closely spaced microphones, also referred to as the spatial coherence, can be approximated by a semi-empirical model, named the Corcos model. Due to the dependency of the Corcos model on the airflow velocity, the measured spatial coherence provides information on the sought quantity. Speed and direction can be resolved by fitting the measured spatial coherence to the analytical Corcos model in the least squares sense. The accuracy of the proposed method is investigated across a range of wind speed between 0.5 and 12 ms⁻¹ and all directions, using observation lengths from 5 s to 1 h. The audio samples under test were recorded indoors and outdoors and labeled by an ultrasonic anemometer. The evaluation results show that the accuracy can be increased by reducing the time resolution.