Abstract
We present a method for tempo estimation from audio recordings based on signal processing and peak tracking, and not depending on training on ground-truth data. First, an accentuation curve, emphasizing the temporal location and accentuation of notes, is based on a detection of bursts of energy localized in time and frequency. This enables the detection of notes in dense polyphonic texture, while ignoring spectral fluctuation produced by vibrato and tremolo. Periodicities in the accentuation curve are detected using an improved version of autocorrelation function. Hierarchical metrical structures, composed of a large set of periodicities in pairwise harmonic relationships, are tracked over time. In this way, the metrical structure can be tracked even if the rhythmical emphasis switches from one metrical level to another. This approach, compared to all the other participants to the Music Information Retrieval Evaluation eXchange (MIREX) Audio Tempo Extraction competition from 2006 to 2018, is the third best one among those that can track tempo variations. While the two best methods are based on machine learning, our method suggests a way to track tempo founded on signal processing and heuristics-based peak tracking. Moreover, the approach offers for the first time a detailed representation of the dynamic evolution of the metrical structure. The method is integrated into MIRtoolbox, a Matlab toolbox freely available.
Highlights
Detecting tempo in music and tracking the evolution of tempo over time is a topic of research in Music Information Retrieval (MIR) that has been extensively studied in recent decades
While the two best methods are based on machine learning, our method suggests a way to track tempo founded on signal processing and heuristics-based peak tracking
To give an indication of metrical activity that would not reduce solely on tempo but takes into consideration the activity on the various metrical levels, we introduce a new measure, called dynamic metrical centroid, which assesses metrical activity based on the computation of the centroid of the periods of a range of selected metrical levels, using their autocorrelation score as weight
Summary
Detecting tempo in music and tracking the evolution of tempo over time is a topic of research in Music Information Retrieval (MIR) that has been extensively studied in recent decades. We present a method that relates to a more classical approach based on signal processing and heuristics-based data extraction. This classical approach generally detects in a first step the temporal repartition of notes, leading to an accentuation curve (or onset detection curve) that is further analyzed, in a second step, for periodicity estimation. Hierarchical metrical structures, composed of a large set of periodicities in pairwise harmonic relationships, are tracked over time in parallel. In this way, the metrical structure can be tracked even if the rhythmical emphasis switches from one metrical level to another. This paper discusses the state of the art in more details, provides a more detailed and accurate description of the proposed method, and provides an extended bibliography of the MIREX Audio Tempo Extraction competition [5]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
