Acoustic source detection and localization using generalized likelihood ratio test in the spherical harmonic domain

Abstract

In recent years, spherical microphone arrays have gained significant attention for analyzing acoustic signals due to their consistent spatial resolution in both elevation and azimuth, added to the ability to transform signals into the spherical harmonics domain. This paper deals with the localization of active sound sources in the presence of noisy and silent recordings. In the literature, the aspects of detection and localization of acoustic signals are often addressed separately. However, the present paper proposes the spherical harmonic generalized likelihood ratio test (SH-GLRT) algorithm which concurrently addresses both aspects while maintaining a fixed false-alarm probability (Pfa). In this context, the acoustic source localization is considered as a detection problem, and the necessary mathematical development is provided. Precisely, the resulting detection thresholds for four spherical harmonics orders are considered, and the detection performance is assessed and compared to the state-of-the-art methods. In addition, the LOCATA challenge dataset is employed for validation, which emphasizes the efficiency of the proposed method.