The invention provides an intelligent housing voice interaction system. A method comprises the following steps: A1, collecting a voice sample by adopting a microphone array, and carrying out signal noise reduction and voice detection; A2, if the voice sample contains voice signals, estimating the number and orientation of a signal source by adopting a 2D_MUSIC algorithm; A3, according to the orientation of the signal source, calculating the weight vector of the signals according to MV_Bearnforning, and carrying out weighting processing, thus forming voice wave beams of the voice sample; A4, carrying out vocal print match with a voice wave beam series saved in the system, and if unsuccessful match exists, adding the voice wave beams with unsuccessful match into a series list; and A5, according to vocal print clustering, regularly aggregating the similar voice wave beams to the same voice signals, and thus the system recognizes the same voice signals as the voice of the same person.