A neural network approach for preclassification in musical chords recognition

Automatic music transcription is a complex task involving signal processing, pattern-matching, signal classification, and the integration of musical knowledge. Existing systems use both bottom-up and top-down approaches. Their performance is determined by the parameters extraction algorithms and the number of different patterns to discriminate. We propose a neural network based preclassification approach to allow a focused search in the chords recognition stage. The specific case of the 6-string standard guitar is considered. The preclassification algorithm outputs the number of strings in a chord and the estimated hand position on the guitar neck.