Abstract: In order to solve the problem of automatic recognition of broadcast, an automatic recognition method of FM radio is proposed based on clustering. Using the radio monitoring equipment to obtain the audio data whose attributes are normal broadcast and black broadcast, the audio data is transformed into text, and these texts are analyzed, including calculate their feature weight, construct the vector space model for text hierarchical clustering, and then extract key words from various types of text as the initial corpus for identifying broadcast attributes and topic categories. During the automatic recognition process, the corpus can be further updated by manually confirming where the broadcast with unknown attribute. The experiment results show that this method can effectively identify the broadcast attributes and categories, and help for the radio monitoring department.