The digital music archive has been enriched with descriptions concerning the musical parameters tempo and pitch. Music Information Retrieval (MIR) techniques have been applied: computer analysis methods that extract musical information from digital audio. The used techniques have been tested extensively and were evaluated positively, but since all the data have been created by automated processes and could not be checked individually because of the size of the archive, the results of the MIR are not infallible.
The tempo analysis has been performed by the software
by Simon Dixon. This software suggests timestamps on which a 'beat' in
the music occurs. From this list of temporal data, the actual tempo
(and related temporal information) could be derived. Also a graph
was made for each song, visualising the tempo development of the
whole piece at first glance.
In general, but also in this case, there must be remarked that tempo annotation can hold a degree of ambiguity: research pointed out that there are plural temporal levels within one piece of music. Manual annotation of tempo enforced this hypothesis of ambiguity. Listeners can attribute several tempi to a piece of music. These tempi are mostly related to each other. For example are the tempi 60 BPM (beats per minute) and 120 BPM closely related. The polyrhythmic character of some African music enforces the ambiguity of tempo annotation, pointing towards triple, quadruple and quintuple relations. Tempo annotation by software faces the same problem: only one tempo can be attributed and this does not necessarily match the tempo of an individual (or your) tempo annotation. Keep in mind that another tempo than expected, can indeed be a correct, though other levelled, tempo! The most often occurring relationship is the halving and doubling of the tempo.
This possible ambiguity aside, we can say that the BeatRoot gives an accurate annotation of the tempo. Random chosen songs were annotated by BeatRoot and its results were compared with manual annotations made by a group of ten people. 67% of the tempi were annotated identically, 26% were annotated in a related tempo and 7% were assigned a non-relating (wrong?) tempo.
The BeatRoot assigns tempi between 60 and 200BPM. From the BeatRoot annotations several temporal characteristics could be derived: tempo (average and median), standard deviation (deviation of general tempo), minimal and maximal tempo, regression (tempo development, percentage of acceleration/deceleration of the beginning tempo). On the website, four can be consulted: average and mean tempo, standard deviation and regression. For every song, there is a graph visualising the tempo development in time: X-axis represents the time (in sec), while Y-axis shows the tempo (BPM). On top of these annotations, a fitting curve has been calculated, showing the regression (green line). The regression is a value above or below zero. A negative value implies that the song in its totality has decelerating tempo, while a positive value indicates a global acceleration. The regression value is the percentage that the tempo increases/decreases per second (average of the entire song). The right side of the graph shows a histogram clustering how many times the tempi did occur. The width of all these annotations makes the standard deviation: a small value implies a stable tempo while a larger value means more rubato or changing tempo.
The 5 graphs visualise characteristics of the recording that are hard to express with
semantic terminology. Sometimes the Western musical concepts lack,
sometimes varying elements over time are easier to visualise in a
graph than in words. The added value for sure is obvious!
May it stimulate further research!