Broadband Music: Opportunities and Challenges for Multiple Source Localization

It is well-known that the subspace MUltiple SIgnal Classification (MUSIC) method provides high-resolution spatial spectral estimates in narrowband signal environments. However, for broadband signals, such high-resolution methods still elude researchers. This paper proposes a broadband version of the MUSIC method using a parameterized version of the spatial correlation matrix. The proposed algorithm utilizes both inter-microphone amplitude and phase differences; as a result, simulation results show that the proposed method allows two broadband sources which are not resolvable by conventional steered beamforming to be accurately resolved.