Skip to Main Content
Various studies on music information retrieval have been conducted. Most of them used low-level features such as spectral and cepstral features, which are effective to some extent but have a limit because they do not directly or clearly correspond to any musical aspects. In this paper, we design new audio features to be extracted from bass lines.Because bass lines play important roles in both harmonic and rhythmic aspects and have a different style for each music genre, our bass-line features are expected to improve the similarity measure and classification accuracy. We also report applications of our bass-line features to music genre classification and musical similarity calculation.