This paper describes a robust algorithm for sound source localization in conference rooms. The method used is a modi¯ed steered response power phase alignment transform algorithm. The results are obtained by processing real data recorded in a typical conference room, and they are compared to data obtained from a simple free-field model. The algorithm demonstrates good accuracy for ¯nding the correct angle of arrival for the dominant speaker in the room and works well for speech sources. The algorithm integrates well with subband decomposition and is suited for realtime applications.