Aeroacoustic Source Separation using RPCA of Microphone Array Signals

We present an application of Robust Principal Component Analysis (RPCA) to acoustic measurements, where the aim is to accurately distinguish between two acoustic sources in noisy signals. Here, we analyze microphone array data from an individual vortex ring (VR) convecting past a semi-infinite half-plane in an anechoic chamber. VR generation is impulsive in character and acts as a second source, producing a weak shock wave. Acoustic pressure measurements using a circular array of 12 microphones centered on the VR/half-plane source are sampled synchronously with high speed Schlieren imaging of VR motion. In this application, RPCA is used to decompose microphone array signals into low-rank and sparse components in an attempt to separate the two sources. Temporal alignment of each acoustic source is performed by steering the array, which encourages higher representation of the desired source's signal energy in the low-rank portion of the RPCA decomposition. RPCA-estimated acoustic source waveforms are then used to estimate sound source parameters for the VR/half-plane interaction.