A multi-stage auditory model for binaural sound localization using the locally competitive algorithm

The human auditory system's ability to accurately localize sounds is essential for navigating and interpreting our environment. However, modern hearing aids and cochlear implants often disrupt critical binaural hearing cues, posing challenges for individuals with hearing impairments. Additionally, precise sound localization is vital for technologies such as smart-home devices, autonomous vehicles, and robotics. In this work we introduce a brain-inspired neural network model, utilizing sparse coding techniques, that can achieve high accuracy in azimuthal sound localization by leveraging binaural and spectral auditory cues. The proposed model achieves a localization accuracy of 95%, comparable to human performance, through a novel application of the Locally Competitive Algorithm (LCA) for efficient sparse coding of auditory signals. This approach not only enhances our understanding of neural auditory processing but also holds promise for improving hearing aids and other auditory-based technologies, thereby advancing individual well-being and technological interactions within complex auditory environments.