"The purpose of this proposal is to generalize the subtractive sound synthesis method for digital audio applications.Whereas the basic idea of Additive Synthesis consists in adding the sinusoidal components of a Fourier series, the analog Subtractive Synthesis aims to imitate a given periodic sound by filtering a fixed excitation signal (usually square, triangle or sawtooth signals). With this source-filter principle, the spectral envelope is reproduced by designing the filter.For some decades, the Additive Synthesis has been significantly improved for computer applications, but the Subtractive Synthesis not. Actually, on one hand, the recent digital subtractive synthesizers focus on modelling old analog synthesizers since the 1960s (from which the Moog is the most popular), and on the other hand, the source-filter principle has been improved for the particular case of the speech coding and synthesis only.Here, we plan to apply recent advances of the audio processing field to derive a new Extended Subtractive Synthesis method. For example, we will study the use of time-varying filters and pitch trajectories to reproduce transients, modulations and roughness. With learning from recorded sounds, we will estimate the parameter trajectories. Moreover, we will use knowledge about psychoacoustics to imitate the original sound from a perceptual viewpoint. This new synthesis technique will have applications in sound synthesis of harmonic musical sounds, and we plan also to synthesize animal sounds. The new method will also be useful for computer game audio. Moreover, most of usages of current Additive Synthesis could be applied, such as time or frequency stretching or noise removal. The main benefits of the Subtractive Synthesis are the reduced number of parameters and the easier control of the resynthesis."