Automatic music transcription using CNN neural networks on segmented audio
Hasan H. Aljamea, Ebrahim A. Mattar · IET conference proceedings. · 2022
Automatic Music Transcription (AMT) converts an audible musical piece into a Music sheet (also known as Music Score). The musical piece is written down in musical notation, instructing musicians how to reproduce the same music piece. Music transcription plays a huge part in preserving Musical culture. However, it is a laborious and time-consuming job to be done manually. Therefore, early attempts on Automatic Music Transcription depend on digital signal processing [1] [2]. The method that will be explored in this paper is by classifying musical notes from Short-time Fourier transform (STFT) images of music piece audio segments using convolutional neural network (CNN), highlighting the limitation of this method.