Residual noise in the aggregate beamformer

The aggregate beamformer samples data from an array of sensors in a random sequence and reconstructs the beamformed signal by time-aligning the data in a 'sequencing buffer'. The reconstructed signal is a highly over-sampled version of the corresponding delay-and-sum beamformer output except for an additional broadband noise component. To remove most of the noise power, while leaving the desired signal intact, the data is low-pass filtered and decimated, leaving a small amount of 'residual noise'. The aggregate beamformer offers reduced hardware complexity by removing the need for anti-alias filters and provides high time-delay resolution without the need for data interpolation or over-sampling on every channel. The total sampling rate is determined by the desired residual noise level and is independent of the number of sensors in the array, making beamforming arrays with many elements, such as for auditoria and other large venues, very practical. The paper describes the aggregate beamformer and demonstrates that modifying the statistical distribution of the channel sampling sequence to shape the noise in the sequencing buffer can further reduce the residual noise.