CNN-based processing of radio frequency signals for augmenting acoustic source localization and enhancement in UAV security applications

A novel source localization algorithm is investigated for application to MAVs. It introduces a DOA and distance prediction method based on processing RF data. A CNN with 2-stage network is proposed to perform this task with the perspective of helping localization and enhancement of acoustic signals for UAV security applications. An experimental sensor data streaming architecture is also presented where the processing complexity is concentrated on a ground station with control capacity. Experiments and results are then discussed.