UAV identification from acoustic signals using statistical learning: A state-of-the-art

Over the last few years, the deployment of Unmanned Aerial Vehicles (UAVs) has sky-rocketed, pushed by the ever-growing range of commercial applications and various malicious purposes. In order to protect sensitive facilities and public areas, their effective and prompt localization and identification is necessary. As no modality (electro-optical, radio, etc.) is yet able to single-handedly perform both of these tasks, a combination of several sensors is required to fulfill these objectives under all real life conditions. Among the different modalities, acoustics relies on pressure signals captured by single microphones or phased arrays to carry out these tasks and has proven its detection efficiency for short distances. Whereas the localization step can be achieved by a set of well-known techniques widely described in the literature, the identification stage has not reached the same level of maturity. Therefore, this paper focuses on the latter and aims at presenting a state-of-the-art of classification techniques based on statistical learning and dedicated to UAV acoustic signatures. Additionally, a specific care is taken to assess the suitability of the reviewed techniques to real life conditions and requirements.