A comprehensive review of sound source localization methods for robotics

Abstract

Sound source localization (SSL) is a key technology in robotics that allows machines to detect and locate auditory cues in real time. This review provides a thorough examination of SSL techniques classified into classical, artificial intelligence (AI), and hybrid methods. Classical methods, which account for 44% of reviewed studies, excel in computational efficiency and reliability under controlled conditions but have limitations in dynamic environments. AI methods, which account for 16% of studies, use deep learning to adapt to complex scenarios, but they require large datasets and computational resources. Hybrid methods, which combine classical signal processing and AI, are the most robust and accurate, with an average accuracy of 97.45%. The review also looks at the role of microphone arrays in SSL performance, revealing that systems with ten or more microphones achieve the highest accuracy of 99.23%, while single- and dual-microphone systems still perform competitively (97.60% and 97.21%, respectively). These findings suggest that hybrid methods combined with larger microphone arrays are the most effective SSL solution in robotics, balancing precision and adaptability. This paper discusses current SSL trends, challenges, and future research directions, providing insights for the development of advanced auditory systems capable of reliable performance in dynamic, real-world environments.