A Blind Source Separation Framework for Ego-Noise Reduction on Multi-Rotor Drones

Abstract

Acoustic sensing from a multi-rotor drone is heavily degraded by the strong ego-noise produced by the rotating motors and propellers. To address this problem, we propose a blind source separation (BSS) framework that extracts a target sound from noisy multi-channel signals captured by a microphone array mounted on a drone. The proposed method addresses the challenging problem of permutation alignment, in extremely low signal-to-noise-ratio scenarios (e.g. SNR <-15 dB), by performing clustering on the time activities of the separated signals across frequencies. Since initialization plays an important role to the success of clustering, we propose a pre-processing algorithm which uses time-frequency spatial filtering (TFS) to generate a reference to pre-align the permutation. The pre-alignment not only improves the performance of clustering and permutation alignment, but also solves the target-channel selection problem for BSS. The proposed method integrates the advantages of both TFS and BSS. Experimental results with real-recorded data show that the proposed method is capable of processing the audio stream continuously in a blockwise manner and also remarkably outperforms the state of the art.

Singular Spectrum Analysis for Source Separation in Drone-Based Audio Recording

“…Wang [9] proposes a BSS frame that extracts a target sound from the noisy multi-channel signals picked up by a set of microphones mounted on a drone. Thus, the frame improves the sound by treating the target and noise signals equally and separating the sources from the mixed signals captured by the microphone array [44].…”

“…Note that since the eigenvectors are unitary vectors, ||q i || = 1 and the projection is simplified to a single inner product. Additionally, the calculus of P can be further simplified into a single matrix product, as detailed in (9).…”

“…In this article, we study how to reduce drone-noise from single-channel recordings. This is seen as a blind source separation (BSS) problem [9]. We thus approach it using the singular spectrum analysis (SSA) algorithm [10] which, in contrast to other methods such as active noise canceling [11], only needs one observation.…”

Applying the majority voting rule in acoustic detection and classification of drones

“…Low signal-to-noise-ratio (SNR) can be a problem in detecting and classifying sound events recorded from a drone, mainly because of ego-noise, the noise produced by the drone. This problem can be addressed using classical signal processing noise reduction algorithms, including frequencyspatial filtering techniques, effective in blind source separation problems [16]. More recently, methods based on Deep Neural Networks (DNN) [17] are also being used to enhance speech signals captured using drones.…”

Uma Plataforma para Redução e Separação de Ruído em Drones Equipados com Sensores Sonoros

“…Um dos objetivosé transmitir o sinal do equipamento de gravação deáudio para o operador do drone. Para resolver o problema de eliminar ou separar o ruído de drone, existem técnicas como Blind Source Separation (BSS) [Wang and Cavallaro 2020] que, por sua vez, faz uso do algoritmo de Singular Spectrum Analysis (SSA).…”