MUSIC-Based Sound Source Localization Algorithm from UVA-Embedded Microphone Array

The application of an unmanned aerial vehicle (UAV) equipped with microphone array for field search and rescue has demonstrated good potentials. One of the main issues in using a multirotor UAV for sound source localization is that the ego noise of the UAV’s rotors interferes with the audio observation and degrades the sound source localization performance. This paper introduces a variant of the Multiple signal classification (MUSIC) technique to audio processing embedded in the UAV scenarios, suppressing background noise by employing the noise datasets to reconstruct the covariance matrix. As simulation results are described, the performance improvement is achieved while using the new approach compared with the generalized cross-correlation with phase transform (GCC-PHAT), non-linear generalized cross-correlation (GCC-NONLIN) and conventional MUSIC method.