Abstract: The essential requirement for noise control is to identify the noise sources accurately. The space directivities of these noises play important roles in the acoustic sources identification and prediction. To improve the space resolution of noise-source maps, the extended acoustic source (which consists of some monopole sources in a certain shape) is used to model the noise-source. The generalized inverse beamforming (GIB) algorithm is performed to discriminate the complex extended acoustic sources. Based on the simulation, the performances of GIB algorithm are discussed, such as the relationship between sound sources identification performance and the number of microphones, the influence of measuring distance, the space resolution of GIB. Then the extended autospectra algorithm and GIB algorithm were compared by the identification of point source, extended sources (multiple monopole sources were closely placed). The result shows that the identification performance of GIB algorithm is not easy to be influenced by the number of microphone array. Compared with autospectra algorithm, the better performance of point sources and extended sources localization was gained.