The what and where of synchronous sound perception

Sound localization and identification are challenging in acoustically rich environments. The relation between these two processes is still poorly understood. As natural sound-sources rarely occur exactly simultaneously, we wondered whether the auditory system could identify ('what') and localize ('where') two spatially separated sounds with synchronous onsets. While listeners typically report hearing a single source at an average location, one study found that both sounds may be accurately localized if listeners are explicitly being told two sources exist. We here tested whether simultaneous source identification (one vs. two) and localization is possible, by letting listeners choose to make either one or two head-orienting saccades to the perceived location(s). Results show that listeners could identify two sounds only when presented on different sides of the head, and that identification accuracy increased with their spatial separation. Notably, listeners were unable to accurately localize either sound, irrespective of whether one or two sounds were identified. Instead, the first (or only) response always landed near the average location, while second responses were unrelated to the targets. We conclude that localization of synchronous sounds in the absence of prior information is impossible. We discuss that the putative cortical 'what' pathway may not transmit relevant information to the 'where' pathway. We examine how a broadband interaural correlation cue could help to correctly identify the presence of two sounds without being able to localize them. We propose that the persistent averaging behavior reveals that the 'where' system intrinsically assumes that synchronous sounds originate from a single source.

Regularization by Denoising for simultaneous source separation

Denoisers can help solve inverse problems via a recently proposed framework known as regularization by denoising (RED). The RED approach defines the regularization term of the inverse problem via explicit denoising engines. Simultaneous source separation techniques, being themselves a combination of inversion and denoising methods, provide a formidable field to explore RED. We investigate the applicability of RED to simultaneous-source data processing and introduce a deblending algorithm named REDeblending (RDB). The formulation permits developing deblending algorithms where the user can select any denoising engine that satisfies RED conditions. Two popular denoisers are tested, but the method is not limited to them: frequency-wavenumber thresholding and singular spectrum analysis. We offer numerical blended data examples to showcase the performance of RDB via numerical experiments.