Musical note recognition

Abstract

This project aims to achieve musical notes recognition on polyphonic musical signals. Onset time application is used throughout the whole musical signal for better recognition of notes. Constant QTransform is applied to transform the musical signal from time to frequency domain. A recognition algorithm, which uses filtering, thresholding and template matching techniques, was developed. The results from the recognition process include detection of musical notes, loudness, notes durations and beat values. This algorithm was tested on five musical pieces which consist of melody and bass cords. The results show 100% detection rate. However, there are limitations to this project. The recognition process can be improved in future work.