Near-field acoustic holography (NAH) is an effective tool for visualizing acoustic sources from pressure measurements made in the near-field of sources using a microphone array. The method involving the Fourier transform and some processing in the frequency-wavenumber domain is suitable for the study of stationary acoustic sources, providing an image of the spatial acoustic field for one frequency. When the behavior of acoustic sources fluctuates in time, NAH may not be used. Unlike time domain holography or transient method, the method proposed in the paper needs no transformation in the frequency domain or any assumption about local stationary properties. It is based on a time formulation of forward sound prediction or backward sound radiation in the time-wavenumber domain. The propagation is described by an analytic impulse response used to define a digital filter. The implementation of one filter in forward propagation and its inverse to recover the acoustic field on the source plane implies by simulations that real-time NAH is viable. Since a numerical filter is used rather than a Fourier transform of the time-signal, the emission on a point of the source may be rebuilt continuously and used for other post-processing applications.

The development of unmanned aerial vehicles (UAVs) opens up a lot of opportunities but also brings some threats. Dealing with these threats is not easy and requires some good techniques. Knowing the location of the threat is essential to deal with an UAV that is displaying disturbing behavior. Many methods exist but can be very limited due to the size of UAVs or due to technological improvements over the years. However, the noise produced by the UAVs is still predominant, so it gives a good opening for the development of acoustic methods. The method presented here takes advantage of a microphone array with a processing based on time domain Delay and Sum Beamforming. In order to obtain a better signal to noise ratio, the UAV’s acoustic signature is taken into account in the processing by using a time-frequency representation of the beamformer’s output. Then, only the content related to this signature is considered to calculate the energy in one direction. This method enables to have a good robustness to noise and to localize an UAV with a poor spectral content or to separate two UAVs with different spectral contents. Simulation results and those of a real flight experiment are reported.

Voice Activity Detection (VAD) and Overlapped Speech Detection (OSD) are key pre-processing tasks for speaker diarization. In the meeting context, it is often easier to capture speech with a distant device. This consideration however leads to severe performance degradation. We study a unified supervised learning framework to solve distant multi-microphone joint VAD and OSD (VAD+OSD). This paper investigates various multi-channel VAD+OSD front-ends that weight and combine incoming channels. We propose three algorithms based on the Self-Attention Channel Combinator (SACC), previously proposed in the literature. Experiments conducted on the AMI meeting corpus exhibit that channel combination approaches bring significant VAD+OSD improvements in the distant speech scenario. Specifically, we explore the use of learned complex combination weights and demonstrate the benefits of such an approach in terms of explainability. Channel combination-based VAD+OSD systems are evaluated on the final back-end task, i.e. speaker diarization, and show significant improvements. Finally, since multi-channel systems are trained given a fixed array configuration, they may fail in generalizing to other array set-ups, e.g. mismatched number of microphones. A channel-number invariant loss is proposed to learn a unique feature representation regardless of the number of available microphones. The evaluation conducted on mismatched array configurations highlights the robustness of this training strategy.

The localization of unmanned aerial vehicles is an important topic due to several threats near sensitive sites. Localization based on their sounds has been a particular point of interest in past studies for many years. It requires the use of a microphone array. The positioning of the various microphones making up an antenna defines the intrinsic directivity of the array. In this study, a genetic algorithm is used to determine the microphone positions that optimize directivity in a focus direction and for a frequency, by favoring the narrowness of the main lobe and the reduction of the secondary lobes. The optimization leads to several antennas with a 3D structure similar to that designed in a previous study. A method estimating the direction of arrival of a drone was also presented in that study making use of its acoustic signature to enhance the signal-to-noise ratio and thus improving the estimations. In this paper, an improvement to the method is proposed for tracking the drone’s trajectory. Measurements were conducted to compare the drone locations given by the first designed antenna and the one optimized by the genetic algorithm. Performance on the direction of arrival found is characterized in terms of mean error, standard deviation and root mean square error relative to the GPS reference onboard the UAV. An experiment with the optimized antenna has also been conducted with the drone at a great distance to the antenna to characterize the maximal distance for possible estimations of the direction of arrival. Results show that the method used for the direction of arrival estimation can give a mean error below 10° in azimuth and 5° in elevation. The maximum distance between the antenna and the drone for which the method is able to give estimations is between 240 and 340 m.