FFT onset detector based on work described in Hainsworth, S. (2003) Techniques for the Automated Analysis of Musical Audio. PhD, University of Cambridge engineering dept.See especially p128. The Hainsworth metric is a modification of the Kullback Liebler distance.
The onset detector has general ability to spot spectral change, so may have some ability to track chord changes aside from obvious transient jolts, but there's no guarantee it won't be confused by frequency modulation artifacts.
Hainsworth metric on it's own gives good results but Foote might be useful in some situations: experimental.
What strength of detection signal from Hainsworth metric to use.
What strength of detection signal from Foote metric to use. The Foote metric is normalised to (0..1).
Threshold hold level for allowing a detection.
If triggered, minimum wait until a further frame can cause another spot (useful to stop multiple detects on heavy signals).