Aeroacoustic Source Separation of Non-Stationary Signals in Time Domain With RPCA
Mitchell J. Swann, Adam S. Nickels, Michael Krane, Jeff R. Harris · 2024
An approach using Robust Principal Component Analysis (RPCA) for time domain acoustic source separation of microphone array signals is presented. Representation of non-stationary, potentially impulsive signals in frequency domain can often be ineffective, necessitating a time domain approach. One such application is the aeroacoustic emission of an individual vortex ring interacting with the edge of a semi-infinite half-plane, where microphone signals are polluted by the presence of a shock wave intrinsic to vortex generation. Both of these sources are impulsive and non-stationary. RPCA successfully performs the desired source separation, enabling characterization of the vortex/edge source. In the present application, RPCA shows improved performance relative to other time domain source separation techniques.