What is digital piano voicing?

Digital piano voicing is an attempt to perform the same operations done by the piano tuner, but by working directly on the digital samples recorded from a real piano. It consists in three stages: analysis, voicing and synthesis. Voicing and synthesis are somewhat imbricated, in the sense that one has to listen to the result and modify some parameters values until the resulting sound gives satisfaction.

First stage: analysis

The first stage consists in analyzing each recorded sample. This task is done by using a combination of Fourier analysis (FFT) and Prony analysis (equivalent to Padé analysis), which is particularly devoted to analysis of signals of the form

a(t)  =  sum  ck exp(-dk t) sin(2 pi fk t + pk) ,

that is, consisting in a sum of overtones with respective amplitude ck, decay rate dk, frequency fk and phase pk.

Second stage: voicing

This second stage is the core of digital voicing. It consists in modifying the different parameters issued from analysis, in order to improve the tone quality and the eveness of the instrument. Parameters  dk, fk and pk are used for adjusting:
Parameters ck give the "color" of the sound, and their evolution from note to note give the eveness of the sound. The voicing attempts to modify them as less as possible, correcting however notes where overtones are too weak or too strong with respect to the global color of the piano. Listen to three different voicings of the same note:
mellow voicing
medium voicing
medium-bright voicing

Third stage: synthesis

In this third stage, all components are re-assembled by using the modified parameters issued from the second stage. The result is a new collection of samples which can then be played on digital instruments.

Here are two exampes showing the result obtained by this digital voicing technique:
original sound voiced sound
original sound voiced sound

Surely you will find that there is still some work to be done ! More examples can be found in section Sound examples.