Time-Frequency Beamforming Adaptation for Micro Aerial Vehicle (MAV) Noise Suppression and Source Localization

This paper presents a modified time-frequency based beamforming technique to use both azimuth and elevation information for signal enhancement of very low Signal-to-Noise Ratio (SNR) multichannel auditory MAV (Micro Aerial Vehicle) recordings. The method identifies acoustic sources embedded in strong harmonic rotor noise by exploiting the time-frequency sparsity of source content. Spatially informed multichannel Wiener filters are applied to the individual recording channels for different array steering directions, and a maximum kurtosis measure is utilized to estimate the direction of arrival. Preliminary results are discussed for a 9-element logarithmic spiral array mounted on a MAV in flight.