Abstract:

This computational model of harmonic chord recognition investigates the perception of harmonic chords by peripheral auditory processes and auditory grouping. The frequency selectivity of the auditory system is modeled using a bank of overlapping band-pass filters and a model of inner hair cell dynamics. The periodicity in each frequency channel is determined by autocorrelation. These periodicities are grouped according to pitch class. By computing the intervals between the high priority pitch classes, the model achieves considerable success in recognizing major, minor, dominant seventh, diminished and augmented chords. Its performance is evaluated on chords from instrument samples as well as from an ensemble recording.