In this paper, we describe a novel audio feature extraction method, which can effectively improve the performance of music identification under noisy circumstances. It is based on a dual box approach that extracts from the sound spectrogram point clusters with significant energy variation. This approach was tested in a song finder application that can identify music from samples recorded by microphone in the presence of dominant noise. A series of experiments show that under noisy circumstances, our system outperforms current state-of-the-art music identification algorithms and provides very good precision, scalability and query efficiency.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.