The invention relates to the field of intelligent application type audio signal processing, in particular to an audio retrieval method utilizing transform domain sparsification and compression dimension reduction. According to the method, an audio fingerprint is used as an audio retrieval feature; according to the technology, in the training stage of audio retrieval, the sample audio data is compressed firstly by utilizing transform domain sparsification and compression dimension reduction, and then the audio fingerprint is extracted from the compressed audio sequence, so that the data volumeof the extracted audio fingerprint is reduced, and the data volume of the sample audio fingerprint library is reduced. In the recognition (retrieval) stage of audio retrieval, the same algorithm is used for compressing the to-be-retrieved audio, and then the to-be-retrieved audio fingerprint is extracted to be matched with the sample audio fingerprint, so that the retrieval efficiency is improvedto a great extent.