STRUCTURAL FINGERPRINT BASED HIERARCHICAL FILTERING IN SONG IDENTIFICATION
Qiang Wang, Zhiyuan Guo, Gang Liu, Guo JunAbstract
Automatic song identification has long been a research focus. In this paper, a novel structural fingerprint based hierarchical filtering method is proposed and it consists of two parts: one is the generation of fingerprint with both long structural information and low collision, and the other is an efficient searching algorithm based on a set of selective 2-level filters. Experiments conducted on a database of 10,000 songs show that our approach is fast enough and can achieve the accuracy of 99.7% on 5 second clips with the SNR at 0db comparable to the state-of-the-art.
Read Submission [1002]