Automatic summarization of MP3 music objects
Chih‐Chin Liu, Pang-Chia Yao · 2004
We propose an approach to summarize MP3 music objects automatically. The MP3 music summary is constructed in three steps. First, coefficients extracted from the output of polyphase filters are used to compute the MP3 features. Based on these features, an MP3 music object can be automatically segmented into a sequence of MP3 phrases. Then, the phrase clustering technique is applied to group similar phrases into a cluster. The phrases in a cluster can be considered as repeating patterns in the MP3 song. Finally, the RP-tree algorithm is applied to find the nontrivial repeating patterns such as chorus, refrains, themes, and motives of the song. Experiments are also performed and analyzed to show the effectiveness of the proposed method.