Source enhancement for unmanned aerial vehicle recording using multi-sensory information

A method to effectively capture desired sound signals from an unmanned aerial vehicle (UAV)-mounted audio recording system by utilising accurate rotor noise power spectral density (PSD) estimations of a UAV is proposed. The method seeks to improve the estimation accuracy and robustness of rotor noise PSD by incorporating UAV rotor characteristics in conjunction with microphone signals. Experiment results show rotor noise PSD estimation accuracy to within 5.5 dB log spectral distortion regardless of the presence of surrounding sound sources, with consistent ~28 dB improvement in signal-to-noise ratio, in particular, reduction of rotor noise from the noisy microphone recordings.