Analysis of Musical Recordings and Expressive Sound Synthesis Event

Prof Su is visiting us as part of growing collaboration between NCKU-one of Taiwan's leading technical universities-and Southampton in the area of music technology, performance studies, sound and vibration research and computer science. He will introduce his own research, which is on the border between musicological performance studies, signal processing and computer science in a short talk. You are welcome to bring your lunch, but if you would like to have a free one there will be drinks and sandwiches.

Abstract

Analysis of Musical Recordings

It would be wonderful if one could learn how master musicians execute their performances from the recordings. It is also helpful if a student can learn her/his problems by recording her/his own playing. In the past, such analysis is more subjectively than objectively. Recently, source separation technologies are applied to audio signals. However, for detail analysis of musical performance, it is usually necessary to study every note in a recording. Therefore, we would like to present the extension from source separation to note separation/tracking. In this talk, acoustic recordings of violin solo, violin sonata and violin concerto are used to demonstrate our currently far from perfect results, especially when a recording contains too many instruments. It is expected that objective analysis and evaluation of musical performing can become popular with the advance of technologies.

Expressive Sound Synthesis

Sound/Music synthesis has been widely used. Methods such as wavetable, FM, spectral modeling, waveguide/physical modeling were proposed in the past few decades. Synthesized music is now everywhere, but it is seldom used in solo. Expression is the most important part in playing musical instrument. Musicians can change their way of playing a note using techniques such as vibrato. This is very common for bowed-string instruments and plucked-string instruments. Unfortunately, it is difficult to simultaneously control both temporal and spectral characteristics of a synthesized note to achieved the desired expression for most synthesis methods. In this talk, we would like to demonstrate a LPC-based synthesis method focusing on producing desired expressions. Guqin music will be played in the talk to show why expression is so important to the playing of instruments. Violin and guqin sounds are used as examples to show the flexibility of our method in controlling characteristics such as pitch, volume, timbre, duration and so on.