Near-Field Acoustic Source Localization and Beamforming in Spherical Harmonics Domain
In this paper, a novel method of robust source localization using MUSIC-Group delay (MUSIC-GD) spectrum computed over spherical harmonic components is described. Our earlier work on the MUSIC-GD spectrum has focused on uniform linear array (ULA) and uniform circular array (UCA) for resolving closely spaced speech sources using minimal number of sensors under reverberant environments. However, this work tries to utilize the advantages of the MUSIC-GD spectrum in a spherical harmonics framework that is computationally simple and more accurate. The MUSIC-GD spectrum for spherical harmonic components is first defined. Its advantages in high resolution DOA estimation are also discussed. Several experiments are conducted for 3-D source localization in reverberant environments and the performance of the MUSIC-GD is compared to other conventional methods. Additional experiments on source tracking are also conducted. The results obtained from the MUSIC-GD computed over spherical arrays are motivating enough to further investigate the method for multiple source tracking in reverberant environments.