Extraction of performance data from great piano performances

Abstract

In the past, many great piano performances were recorded in the form of analog which would then be stored in tapes or records. Over the years, these valuable recordings have degraded vastly. The outstanding music pieces by renowned pianists may be lost and forgotten if no attempts were made to recover and reproduce them. The main objective of the project is to extract performance data of great piano performances in the early days and reproduce them with some degrees of quality and enhancement. While it is impossible to reproduce the exact copy of the originals, we can try to preserve as much as possible using modern technology available at present. The project also investigates the methods in which application software could parse sound at a level similar to our human ear perception. MatLab is extensively used as a coding tool in this project. Parameters such as the pitch, duration and volume (or loudness) of music pieces will be extracted and translated to audio formats via conversion between wave file (WAV) and Musical Instrument Digital Interface (MIDI). The outputs show that music information lost due to sound degradation can be restored through investigation and experimentation of signal processing techniques. In addition, musically fulfilling pieces can be reproduced with the aid of up-to-date offline and online resources. These have contributed greatly to the audio quality of the piano performances. Although there is still much room for improvement, this project can be potential for commercial exploitation.