In this paper, a novel audio retrieval method based on fuzzy classification in MPEG-1 (Moving Pictures Experts Group 1) compressed domain is proposed to search information in steaming media on Internet. In order to avoid the huge decoding time cost, scalefactors are directly extracted from MPEG-1 audio frame as features. At the same time, a distance-based fuzzy classification method is put forward to solve the serious misclassification using traditional audio classification. Experimental results show that the recall rate and the precision of the proposed method are higher than the traditional audio retrieval method. And the computational time is dramatically reduced.