Simulation of Human voice Timbre by Orchestration of acoustic Music Instruments

This paper describes a method which simulates the timbre of the human voice through the orchestration of instrumental ensembles for classical music. The spectral envelope of speech is used as a model for the orchestration. The idea of this method is to synthesise the spectral envelope of a phoneme using the spectral envelopes of different sounds of music instruments. Large instrumental databases with standardised intensities like the Virtual Orchestra are required. For each sound of the database, an average spectral envelope is calculated. An error minimisation algorithm optimises the similarity of the orchestration’s envelope in relation to the phoneme. Among other pieces, this method was applied in the orchestra piece Nicanor, premiered 1999 in Stuttgart. A considerable similarity of the respective orchestral sequence to the sound of the whispered word is perceived.