Acoustic source localization and signal separation based on a hybrid circular array

A circular microphone arrays (CMA) with a vertical logarithmic-spacing linear array (LLA) extension is proposed in this paper for localizing sources and separating signals emanated from those sources. A one-stage approach and a two-stage approach are suggested. In the one-stage method, localization and separation are accomplished at one shot. An underdetermined system is formulated and solved by using compressive sensing (CS) techniques such as convex optimization, the focal underdetermined system solver (FOCUSS), and several ad hoc iterative CS algorithms. In the two-stage method, however, localization is first carried out by minimum power distortionless response (MPDR) algorithm and then followed by separation using Tikhonov regularization (TIKR), convex optimization, and iterative methods on the basis of the estimated source bearings. To validate the proposed hybrid array and the localization and separation methods, a three-dimensionally printed circular array comprised of a 24-element CMA and an 8-element LLA is constructed. Objective Perceptual Evaluation of Speech Quality (PESQ) test and a subjective listening test are also undertaken.