A method is described for making natural sounding modifications of the dynamic level of tones produced by acoustic instruments. Each tone is first analyzed in the frequency domain and divided into a harmonic and a noise component. The two components are modified separately using filters based on spectral envelopes extracted from recordings of isolated tones played at different dynamic levels. When transforming from low to high dynamics, additional high frequency partials are added to the spectrum to enhance the brightness of the sound. Finally, the two modified components are summed and a time domain signal is synthesized.

Download