Synthesis of Emotional Expression in Piano Music Today's musicians frequently play sequencers, synthesisers and samplers. Such instruments can be played also by computers, but then a dull, mechanical performance results. A better result can be obtained if a generative performance grammar is applied. The first part of the seminar will give a brief overview of the on-going research project on automatic music performance, started in 1977 at the Speech Music and Hearing Dep. (TMH) at the KTH, Stockholm. The study of the mechanisms involved in the performing action brought to the design of software tools for the simulation of a human player. Different techniques were used in the development of these tools, from expert systems to artificial neural networks (ANNs). In its second part the seminar will be focused on our most recent experimental study, in which different performances of music scores were synthesised so as to elicit listeners' associations to six different emotions (happiness, sadness, tenderness, solemnity, fear, anger) and to a neutral version (no-expression). Synthesised performances were derived from the Director Musices (DM), a software system for automatic music performance developed at TMH. DM was set up according to findings from previous studies in the field of emotional aspects of music performance, carried out by the Psychology of Music group at the Psychology Dep., Uppsala University. A listening test was carried out where twenty listeners were asked to classify the performances according to seven emotional alternatives. Results from the test show that listeners tended to consistently classify the performances in accordance with the intended emotions. During the seminar, it will be asked to the participants to take part to a similar test. back | abstract in Swedish | tables of results | sound examples © Roberto Bresin - Last update: 1999.07.12