The University of Southampton
News

Keyboard magic, from concert pianos to computers

Published: 
24 August 2004

Researchers at the universities of Southampton and Vienna have demonstrated that the complex and individual performance styles of concert pianists such as Glenn Gould and Vladimir Horowitz can be modelled in unique 'performance alphabets', providing a method of recognizing their performance styles by computer, and also, possibly, reconstructing them.

Concert pianist Glenn Gould had a unique and instantly recognizable performance style, for which he is rightly renowned. Indeed, the extent to which pianists such as Gould, Horowitz and Uchida have a discernibly individual style of playing is recognized not just by classical music aficionados, who can hear the differences, but also by computers, which can analyse the differences and model them.

Now a group of researchers at the Universities of Vienna and Southampton have made significant advances in demonstrating repeated and identifiable differences in individual performances of the same work, to such an extent that a 'performance alphabet' could be drawn up for each composer.

Not only does this provide a means of identifying individual pianists using only minimal information from audio recordings, it also means that one can countenance the possibility of modelling aspects of a performance of a piece as it would have been played, for example, by Horowitz.

The researchers at the Medical University of Vienna, led by Professor Gerhard Widmer took two measures of the way the same Mozart sonatas were played by Glenn Gould, Daniel Barenboim, Andras Schiff, Mitsuko Uchida, Roland Batik, and Maria Joao Pires. These measures were tempo, as measured against a fixed tempo, and volume, or loudness. Professor John Shawe-Taylor of the University of Southampton School of Electronics and Computer Science has led a group developing new methods of analysing the results obtained by the Vienna team.

"Different players have different ways of building tension or expression in the music," says John Shawe-Taylor describing the work of the Vienna team, "and they represent this raw data for every note and progression of the music as a trajectory, which can be represented visually in tempo-loudness space as a "performance worm"." The Vienna researchers have constructed a visual representation of these changes, which can also be compared.

The researchers then obtained certain characteristics of the performance by analysing the movement of the worm. "A performer may consistently produce loudness/tempo changes unique to themselves at specific points in a piece," says John Shawe-Taylor, "for example in association with particular cadences."

The researchers at Vienna have for example observed that Mitsuko Uchida demonstrates a particular way of combining a crescendo-decrescendo with a slowing down during a loudness maximum. These patterns were often repeated in Mozart performances by Mitsuko Uchida, but were rarely found when analysing the recordings of other performers.

The novel analysis techniques applied to the performance worm data at the University of Southampton were able to distinguish the different performers based on a relatively small sample of their performances.

On a more speculative note, John Shawe-Taylor says: "Basically we are seeking common patterns across two different ways of looking at an event. On the one hand we have the musical score, and on the other, its interpretation by an individual concert pianist.

"If we could combine our description of the way the piece is performed and the musical score, and discern the connection between the two, then we might be able to generate aspects of a Horowitz performance of a piece that he had never actually played."

Notes for editors

  1. The paper 'Using String Kernels to Identify Famous Performers from their Playing Style' was published in Proceedings of The 15th European Conference on Machine Learning (ECML) and the 8th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD), Pisa, Italy, and received the award for best conference paper. It is available on:http://eprints.ecs.soton.ac.uk/9522/
  2. The research was in part supported by the Engineering and Physical Sciences Research Council, the IST Programme of the European Community, and by the Austrian Fonds zuer forderung der Wissenschaftlichen Forschung.
  3. The University of Southampton is a leading UK teaching and research institution with a global reputation for leading-edge research and scholarship. The University has over 19,200 students and 4800 staff. Its annual turnover is in the region of £250 million.

Articles that may also interest you

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×