Spectrally correlated sound source localization in a noisy envirnoment based on a new mask for DUET
Ali Parsayan, Seyed Mohammad Ahadi · 2011
This paper investigates the use of a new mask for the degenerate unmixing estimation technique (DUET) based on autocorrelation function (ACF) for spectrally correlated sound source localization (SCSSL), such as a helicopter, in time-frequency domain (TFD). Helicopter is a good example of spectrally correlated sound sources since the main and tail rotors of a helicopter are simultaneously turning by a single engine. Rotating speed of both rotors (and therefore rotor noises) are correlated which leads to correlated frequency components in TFD. The ACF matrix of spectrally correlated sound sources in a noisy environment (underdetermined case which sensors count is less than sound sources count) is sparse and unique and can be used to classify and localize such sources using DUET with a new time-frequency domain autocorrelation function matrix mask (TFD-ACF mask). Simulation results for localization of a helicopter in different simulated maneuvers in presence of different environmental background noises confirm the efficiency of our proposed approach in comparison with previous localization methods in underdetermined case.