Musical Timber Assessment Using MFCCs and Total Harmonic Distortion in Matlab

Colour of music known as timber is a significant feature of sound which represents the character of it. If a musical instrument is perceived differently from other type while playing the same note, it is all because of their timber. The most important impact takes place when the music is recorded and the colour is changed. When someone listens to a piece of music he/she can identify whether the music is live or recorded. So if recording procedure and environment recording are set up in a way that they could produce the best timber, then it can be claimed that the timber of music is natural after recording. Timber extraction using Mel Frequency Cepstral Coefficients (MFCCs) method is done in this project to show that visual extraction can do nothing with differentiations between timbers and it is not possible to comment if the music is natural enough after recording or not by only looking at the timber. Then Total Harmonic Distortion calculation is employed in MFCCs to see how timber is distorted after recording, and then a group of normal and expert musicians are asked to comment about those pieces of music to make a comparison and reach an approach for musical timber assessment. Finally an estimation of timber checking is suggested which can be used by sound engineers to check if the recording environment and the musical recording equipment would result in natural timber after recording.