Distinguishing pianos: The difference in similarity

Notating a music piece is not a trivial task. It requires training and experience. This is challenging for new and inexperienced musicians. Automated transcription systems can be very useful in such cases. A piece generally consists of multiple instruments and it is essential to separate them prior to transcription. The challenge aggravates even more when multiple modulations or varieties of the same instrument are present. This is common nowadays and hence, it is essential to distinguish different varieties of the same instrument prior to transcription. In this paper, a line spectral frequency-based technique is used. A highest accuracy of 97.06% was obtained for a dataset composed of 6 different pianos.