Addressing Tempo Estimation Octave Errors in Electronic Music by Incorporating Style Information Extracted from Wikipedia

A frequently occurring problem of state-of-the-art tempo estimation algorithms is that the predicted tempo for a piece of music is a whole-number multiple or fraction of the tempo as perceived by humans (tempo octave errors). While often this is simply caused by shortcomings of the used algorithms, in certain cases, this problem can be attributed to the fact that the actual number of beats per minute (BPM) within a piece is not a listener’s only criterion to consider it being “fast” or “slow”. Indeed, it can be argued that the perceived style of music sets an expectation of tempo and therefore influences its perception. In this paper, we address the issue of tempo octave errors in the context of electronic music styles. We propose to incorporate stylistic information by means of probability density functions that represent tempo expectations for the individual music styles. In combination with a style classifier those probability density functions are used to choose the most probable BPM estimate for a sample. Our evaluation shows a considerable improvement of tempo estimation accuracy on the test dataset.