Microphone-Array Ego-Noise Reduction Algorithms for Auditory Micro Aerial Vehicles

Abstract

When a micro aerial vehicle (MAV) captures sounds emitted by a ground or aerial source, its motors and propellers are much closer to the microphone(s) than the sound source, thus leading to extremely low signal-to-noise ratios (SNR), e.g.-15 dB. While microphone-array techniques have been investigated intensively, their application to MAV-based ego-noise reduction has been rarely reported in the literature. To fill this gap, we implement and compare three types of microphonearray algorithms to enhance the target sound captured by an MAV. These algorithms include a newly emerged technique, time-frequency spatial filtering, and two well-known techniques, beamforming and blind source separation. In particular, based on the observation that the target sound and the ego-noise usually have concentrated energy at sparsely isolated timefrequency bins, we propose to use the time-frequency processing approach, which formulates a spatial filter that can enhance a target direction based on local direction of arrival estimates at individual time-frequency bins. By exploiting the time-frequency sparsity of the acoustic signal, this spatial filter works robustly for sound enhancement in the presence of strong ego-noise. We analyze in details the three techniques and conduct a comparative evaluation with real-recorded MAV sounds. Experimental results show the superiority of blind source separation and timefrequency filtering in low-SNR scenarios.

A semblance based TDOA algorithm for sound source localization

“…Despite the performance with the audio 3, the obtained results showed that the proposed method presents a good performance, making able to retrieve the source location (within an error of 10°) in cases where the signal-to-noise ratio (SNR) was of -16 dB. This is aligned with the results obtained in [10], where methods of state-of-art had their performance reported.…”

DREGON: Dataset and Methods for UAV-Embedded Sound Source Localization

“…Additionally, since the microphones are mounted on the drone itself, they are very close to the noise sources leading to high noise levels. Because of this, the SNR can easily reach -15 dB or less [11] making SSL very difficult. Another factor impacting localization performance is wind noise.…”

“…On the other hand, SSL is a long standing and extensively studied topics in robotics [18]. However, both ego-noise reduction and SSL for the specific setting of UAV-embedded microphones are still rather new topics [3], [19], [4], [14], [20], [5], [11], [11]. Many SSL approaches developed in recent years for robotics are different variations of the MUltiple SIgnal Classification (MUSIC) algorithm.…”

“…To overcome the large training data requirements of DNNs, a partially shared network learning multiple tasks at the same time is implemented. The authors of [11] compare beamforming, blind source separation (BSS) and time-frequency filtering algorithms on UAV-embedded recordings. The recordings are made with a circular microphone array placed over a UAV fixed on a tripod.…”

SVR-CMT Algorithm for Null Broadening and Sidelobe Control

“…Beamforming technique is an important part of array signal processing. It is widely used in wireless communications [1], microphone array signal processing [2] and radar [3]. A typical, representative beamformer, known as minimum variance distortionless response (MVDR) or Capon beamformer [4], minimizes array output power and maintains a distortionless mainlobe response toward the desired signal.…”