In professional performances of piano music the melody voice tends to anticipate the accompaniment by 20-30 milliseconds. This may be due to dynamic differentiation. The melody voice is usually played louder than the accompaniment which may also cause the melody to lead, since the tone production in a grand piano is based on hammers being accelerated towards strings. A hard key strike results in a high hammer velocity whereas a soft key strike results in a low hammer velocity. Consequently, the hammers will arrive at the strings asynchronously if the keys are depressed with varying force. Presuming that the player strives at playing synchronously, this could easily be modeled and hopefully, provide more natural performances of piano music played by computers.
A model of this hypothesis, based on earlier studies, was implemented in the pDM program for real-time musical expression. The model was then tested in a listening test with experienced pianists. The findings indicate that the model simulates the typical asynchronous onsets found in piano music relatively well.